Active sources: 7. Sources represented in this queue: 6. The cron runs daily at 06:00 server time; arxiv RSS is often empty on weekends.
Linked to your results
- score 100arxiv cs.CL (NLP)arxiv:2605.20684unread
Beyond Semantic Similarity: A Two-Phase Non-Parametric Retrieval Workflow for Corporate Credit Underwriting
Linus Ng Junjia, Ezekiel Tee Kongquan, Kelvin Heng, Kenneth Zhu Ke, Zhao Jing Yuan · 2026-05-21
arXiv:2605. 20684v1 Announce Type: new Abstract: Corporate credit underwriting requires analysts to extract actionable evidence from long, heterogeneous financial documents spanning hundreds of pages and multiple languages.
Read next because Beyond Semantic Similarity: A Two-Phase Non-Parametric Retrieval Workflow for Corporate Credit Underwriting overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, under, eval, line, rate, extraction, control, candidates. Source: arxiv cs.CL (NLP).
arXiv:2605.20684v1 Announce Type: new Abstract: Corporate credit underwriting requires analysts to extract actionable evidence from long, heterogeneous financial documents spanning hundreds of pages and multiple languages. Standard Retrieval-Augmented Generation (RAG) pipelines optimize for semantic similarity, which frequently surfaces passages that are topically related but lack decision utility, a problem we term the similarity-utility gap. We propose a two-phase non-parametric retrieval architecture that separates high-recall candidate retrieval from high-precision utility ranking. The first phase combines lexical and dense multilingual retrieval to construct a broad candidate pool. The second phase applies an adaptive retrieval controller that filters candidates using query intent and document structure signals, followed by an LLM-as-a-Judge utility scoring mechanism that ranks passages by analytical usefulness rather than semantic proximity. A context-aware extraction module preserves structural fidelity across narrative text and complex financial tables. The system is deployed entirely on-premise to satisfy enterprise data governance requirements. Evaluated on a multilingual corpus of proprietary financial documents with analyst-curated relevance labels, the system significantly outperforms naive retrieval baselines. In production deployment across more than 800 credit analysts, document review time was reduced from several hours to approximately three minutes, demonstrating the practical value of utility-aware RAG architectures for document-intensive decision-support workflows.
- score 100arxiv cs.CL (NLP)arxiv:2605.20628unread
Divide-Prompt-Refine: a Training-Free, Structure-Aware Framework for Biomedical Abstract Generation
Sylvey Lin, Joe Menke, Shufan Ming, Dongin Nam, Neil Smalheiser, Halil Kilicoglu · 2026-05-21
arXiv:2605. 20628v1 Announce Type: new Abstract: Biomedical abstracts play a critical role in downstream NLP applications, such as information retrieval, biocuration, and biomedical knowledge discovery.
Read next because Divide-Prompt-Refine: a Training-Free, Structure-Aware Framework for Biomedical Abstract Generation overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, strong, text, under, alignment, eval, source, line. Source: arxiv cs.CL (NLP).
arXiv:2605.20628v1 Announce Type: new Abstract: Biomedical abstracts play a critical role in downstream NLP applications, such as information retrieval, biocuration, and biomedical knowledge discovery. However, a non-trivial number of biomedical articles do not have abstracts, diminishing the utility of these articles for downstream tasks. We propose DPR-BAG (Divide, Prompt, and Refine for Biomedical Abstract Generation), a training-free, zero-shot framework that generates coherent and factually grounded abstracts for biomedical articles with full text but no abstract. DPR-BAG decomposes full-text documents into structured rhetorical facets following the Background-Objective-Methods-Results-Conclusions (BOMRC) schema, performs parallel LLM-based summarization for each facet, and applies a final refinement stage to restore global discourse coherence. On PMC-MAD, a distribution-aligned dataset of 46,309 biomedical articles, DPR-BAG improves abstractive novelty over strong extractive and fine-tuned baselines, while maintaining factual consistency. Our ablation study reveals a counterintuitive finding: increasing prompt complexity or explicitly injecting entity-level guidance can degrade factual alignment, highlighting the importance of controlled prompting strategies. These findings underscore the potential of training-free, structure-aware frameworks for scalable biomedical abstract generation in low-resource settings. Our data and code are available at https://huggingface.co/datasets/pmc-mad/PMC-MAD and https://github.com/ScienceNLP-Lab/MultiTagger-v2/tree/main/DPR-BAG.
- score 100arxiv cs.CL (NLP)arxiv:2605.20616unread
Auto-Dreamer: Learning Offline Memory Consolidation for Language Agents
Chongrui Ye, Yuxiang Liu, Yu Wang, Haofei Yu, Yining Zhao, Ge Liu, Julian McAuley, Jiaxuan You · 2026-05-21
arXiv:2605. 20616v1 Announce Type: new Abstract: Language agents increasingly operate over streams of related tasks, yet existing memory systems struggle to convert accumulated experience into reusable knowledge.
Read next because Auto-Dreamer: Learning Offline Memory Consolidation for Language Agents overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, eval, source, line, rate, without, alone, trained. Source: arxiv cs.CL (NLP).
arXiv:2605.20616v1 Announce Type: new Abstract: Language agents increasingly operate over streams of related tasks, yet existing memory systems struggle to convert accumulated experience into reusable knowledge. Retrieval-augmented and structured memory methods record per-session observations effectively, but often couple acquisition and consolidation into a single online process, leaving the agent without a global view across sessions to discover recurring patterns, abstract shared procedures, or prune redundant entries. Inspired by complementary learning systems theory, we propose Auto-Dreamer, a learned offline consolidator for language-agent memory. Auto-Dreamer decouples fast per-session memory acquisition from slow cross-session consolidation. Given a selected working region of a typed memory bank, the consolidator treats the region as read-only evidence, performs bounded tool-use to inspect entries and provenance-linked source trajectories, and synthesizes a fresh compact replacement set that abstracts across sessions and supersedes the original region. We train Auto-Dreamer via GRPO, using end-to-end agent performance as the reward signal to learn how to consolidate memories acquired through fast online experience. Trained on ScienceWorld trajectories alone, Auto-Dreamer outperforms fixed, RL-trained, and prompted memory baselines on ScienceWorld by 7 points while using an active memory bank 12$\times$ smaller than the strongest baseline, and continues to lead on held-out ALFWorld and WebArena without retraining -- using 6$\times$ less memory than the strongest baseline on ALFWorld.
- score 100arxiv cs.CL (NLP)arxiv:2605.20613unread
HRM-Text: Efficient Pretraining Beyond Scaling
Guan Wang, Changling Liu, Chenyu Wang, Cai Zhou, Yuhao Sun, Yifei Wu, Shuai Zhen, Luca Scimeca, Yasin Abbasi Yadkori · 2026-05-21
arXiv:2605. 20613v1 Announce Type: new Abstract: The current pretraining paradigm for large language models relies on massive compute and internet-scale raw text, creating a significant barrier to foundational research.
Read next because HRM-Text: Efficient Pretraining Beyond Scaling overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, arc-c, prefix, token, line, rate, trained, completion. Source: arxiv cs.CL (NLP).
arXiv:2605.20613v1 Announce Type: new Abstract: The current pretraining paradigm for large language models relies on massive compute and internet-scale raw text, creating a significant barrier to foundational research. In contrast, biological systems demonstrate highly sample-efficient learning through multi-timescale processing, such as the functional organization of the frontoparietal loop. Taking this as inspiration, we introduce HRM-Text, which replaces standard Transformers with a Hierarchical Recurrent Model (HRM) that decouples computation into slow-evolving strategic and fast-evolving execution layers. To stabilize this deep recurrence for language modeling, we introduce MagicNorm and warmup deep credit assignment. Furthermore, instead of standard raw-text pretraining, we train exclusively on instruction-response pairs using a task-completion objective and PrefixLM masking. Serving as an empirical existence proof of efficient pretraining, a 1B-parameter HRM-Text model trained from scratch on only 40 billion unique tokens and $1,500 budget achieves 60.7% on MMLU, 81.9% on ARC-C, 82.2% on DROP, 84.5% on GSM8K, and 56.2% on MATH. Despite utilizing roughly 100-900x fewer training tokens and 96-432x less estimated compute than standard baselines, HRM-Text performs competitively with 2-7B parameter open models. These results demonstrate that co-designing architectures and objectives can radically reduce the compute-to-performance ratio, making pretraining from scratch accessible to the broader research community.
- score 100arxiv cs.CL (NLP)arxiv:2605.20602unread
Self-Training Doesn't Flatten Language -- It Restructures It: Surface Markers Amplify While Deep Syntax Dies
Ming Liu · 2026-05-21
arXiv:2605. 20602v1 Announce Type: new Abstract: Successive self-training on a language model's own outputs is widely characterized as a process of flattening: diversity drops, distributions narrow, and the text becomes "more like itself.
Read next because Self-Training Doesn't Flatten Language -- It Restructures It: Surface Markers Amplify While Deep Syntax Dies overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: marker, text, word, rect, under, rate, control, does. Source: arxiv cs.CL (NLP).
arXiv:2605.20602v1 Announce Type: new Abstract: Successive self-training on a language model's own outputs is widely characterized as a process of flattening: diversity drops, distributions narrow, and the text becomes "more like itself." We provide evidence that this characterization is incomplete. Across eleven generations of self-training on five models (GPT-2 124M, Pythia-410M, Pythia-1.4B, OPT-1.3B, Pythia-2.8B), language is not flattened uniformly -- it is restructured. Surface markers (discourse connectives, hedges, em-dashes) rise, while mid- and deep-syntactic structures (questions, parentheticals, passives, subjunctives) collapse. We formalize this asymmetric collapse as the Structural Depth Hypothesis (SDH): the per-generation decay rate of a linguistic feature is predicted primarily by its structural depth -- the number of nested syntactic dependencies it requires -- and only secondarily by its generation-zero output frequency. Pooling 17-feature panels from five models spanning three architecture families (N=85), the pooled Spearman correlation is rho=0.540 (p < 10^{-6}; cluster-bootstrap 95% CI [0.434, 0.634]), while frequency is a substantially weaker predictor (rho=0.225). A matched human-text fine-tuning control yields rho=0.039 (p=0.88), confirming the gradient is self-training-specific. We further document a Superficial Complexity Paradox: aggregate complexity proxies (dep-tree depth, TTR, word length) all rise as the underlying clause structure dies, with direct implications for training-data curation and LLM-text detection.
- score 100arxiv cs.CL (NLP)arxiv:2605.20588unread
Direct Translation between Sign Languages
Zetian Wu, Bowen Xie, Wuyang Meng, Milan Gautam, Stefan Lee, Liang Huang · 2026-05-21
arXiv:2605. 20588v1 Announce Type: new Abstract: The field of sign language translation has witnessed significant progress in the translation between sign and spoken languages, but the translation between sign languages remains largely unexplored and out of reach.
Read next because Direct Translation between Sign Languages overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, rect, line, rate, without, language, model. Source: arxiv cs.CL (NLP).
arXiv:2605.20588v1 Announce Type: new Abstract: The field of sign language translation has witnessed significant progress in the translation between sign and spoken languages, but the translation between sign languages remains largely unexplored and out of reach. The latter can help 1.5 billion deaf and hard-of-hearing (DHH) people worldwide communicate across language barriers without relying on hearing interpreters or written-language fluency. The cascade approach composing separate sign-to-text, text-to-text, and text-to-sign systems suffers from error propagation and extra latency as well as the loss of information unique in the visual modality. We aim to develop direct sign-to-sign translation. However, a large-scale open-domain parallel corpus has not been curated between sign languages. To enable direct translation between sign language utterances, we use back-translation to produce synthetic sign-sign pairs from unaligned individual language utterance-sign corpora. Using this data, we jointly train a single MBART-based model for both text->sign (T2S) and sign->sign (S2S). On synthetically generated paired sets between American Sign Language (ASL), Chinese Sign Language (CSL), and German Sign Language (DGS), our direct S2S method outperforms the cascaded baseline on geometric sign error metrics (20% lower DTW-aligned MPJPE) and language matching metrics after predicted sign utterances are translated back to sentences (50% high BLEU-4) while achieving a roughly 2.3* speedup. On a small set of pre-existing cross-lingual sign data, we find similar improvements for our proposed method.
- score 100arxiv cs.CL (NLP)arxiv:2605.20529unread
Collocational bootstrapping: A hypothesis about the learning of subject-verb agreement in humans and neural networks
Claire Hobbs, R. Thomas McCoy · 2026-05-21
arXiv:2605. 20529v1 Announce Type: new Abstract: In what ways might statistical signals in linguistic input assist with the acquisition of syntax?
Read next because Collocational bootstrapping: A hypothesis about the learning of subject-verb agreement in humans and neural networks overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)". Matching terms: word, rect, rate, language. Source: arxiv cs.CL (NLP).
arXiv:2605.20529v1 Announce Type: new Abstract: In what ways might statistical signals in linguistic input assist with the acquisition of syntax? Here we hypothesize a mechanism called collocational bootstrapping, in which regularities in word co-occurrence patterns can provide cues to syntactic dependencies. We investigate whether this mechanism can support the acquisition of English subject-verb agreement. First, we simulate language acquisition by training neural networks on synthetic datasets that vary in how predictable their subject-verb pairings are. We find that there is a range of variability levels at which these statistical learners robustly learn subject-verb agreement. We then analyze the variability of subject-verb pairings in child-directed language, and we find that the variability in such data falls within the range that supported robust generalization in our computational simulations. Taken together, these results suggest that collocational bootstrapping is a viable learning strategy for the type of input that children receive.
- score 100arxiv cs.CL (NLP)arxiv:2605.20201unread
Long-Context Reasoning Through Proxy-Based Chain-of-Thought Tuning
Miao Li, Irina Saparina, Alexander Gurung, Mirella Lapata · 2026-05-21
arXiv:2605. 20201v1 Announce Type: new Abstract: Recent large language models support inputs of up to 10 million tokens, yet they perform poorly on long-context tasks that require complex reasoning.
Read next because Long-Context Reasoning Through Proxy-Based Chain-of-Thought Tuning overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, text, under, token, line, rate, full, chain. Source: arxiv cs.CL (NLP).
arXiv:2605.20201v1 Announce Type: new Abstract: Recent large language models support inputs of up to 10 million tokens, yet they perform poorly on long-context tasks that require complex reasoning. Such tasks can be solved using only a subset of the input -- a proxy context -- rather than the full sequence. Despite sharing the same underlying reasoning process, models exhibit a significant performance disparity between proxy and full contexts. To improve long-context reasoning, we propose ProxyCoT, a novel training framework that transfers reasoning capabilities from short proxy contexts to full long contexts. Specifically, we first obtain high-quality chain-of-thought reasoning traces on proxy contexts through reinforcement learning or distillation from a larger teacher model, and then ground the generated traces in full long contexts with supervised fine-tuning. Experiments across different datasets demonstrate that ProxyCoT consistently outperforms strong baselines with reduced computational overhead. Furthermore, models trained with ProxyCoT generalize their long-context reasoning capabilities to out-of-domain tasks.
- score 100arxiv cs.CL (NLP)arxiv:2605.20199unread
FlowLM: Few-Step Language Modeling via Diffusion-to-Flow Adaptation
Runzhe Zhang, Letian Chen, Wenpeng Zhang, Zhouhan Lin, Peilin Zhao · 2026-05-21
arXiv:2605. 20199v1 Announce Type: new Abstract: We present FlowLM, a flow matching language model transformed from pre-trained diffusion language models via efficient fine-tuning.
Read next because FlowLM: Few-Step Language Modeling via Diffusion-to-Flow Adaptation overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, epochs, line, rate, trained, language, model. Source: arxiv cs.CL (NLP).
arXiv:2605.20199v1 Announce Type: new Abstract: We present FlowLM, a flow matching language model transformed from pre-trained diffusion language models via efficient fine-tuning. By re-aligning the curved sampling trajectories of diffusion models into straight-line flows, FlowLM enables high quality few-step generation that rivals or even outperforms the quality of 2,000-step diffusion sampling with very few training epochs. Remarkably, finetuned FlowLM reaches performance saturation with only half as many training epochs as training from scratch, both approaches greatly outperforming the original diffusion model, thereby validating our method. Furthermore, we validate a more effective training objective for flow matching: predicting clean data to consistently guide the sampling process towards the true data distribution. Empirical results demonstrate that our approach is highly effective for high-quality, few-step text generation.
- score 100arxiv cs.CL (NLP)arxiv:2605.20196unread
Data Scaling as Progressive Coverage of a Predictive Contribution Spectrum
Zihui Song, Shihao Ji, Hongxi Li, Shuaizhi Cheng, Chunlin Huang · 2026-05-21
arXiv:2605. 20196v1 Announce Type: new Abstract: We investigate the hypothesis that real-data scaling laws are governed by progressive coverage of a latent predictive contribution spectrum rather than by token-frequency tails alone.
Read next because Data Scaling as Progressive Coverage of a Predictive Contribution Spectrum overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, text, token, line, alone, suffix. Source: arxiv cs.CL (NLP).
arXiv:2605.20196v1 Announce Type: new Abstract: We investigate the hypothesis that real-data scaling laws are governed by progressive coverage of a latent predictive contribution spectrum rather than by token-frequency tails alone. We work with a suffix-automaton representation of text corpora and define a data-intrinsic global-KL predictive contribution spectrum, in which each state contributes according to its empirical mass times its KL deviation from a global next-token baseline. Across 12 real corpora, the tail slope of this spectrum is already strongly correlated with the empirical data-scaling exponent of a fixed small GPT learner. We then go beyond slope correlation and define, for each training size N, an effective truncation rank K(N) by matching the observed excess loss to the residual tail mass of the prepared 1000k global-KL spectrum. Empirically, log K is close to linear in log N, with pooled R^2 about 0.96 for the raw spectrum and R^2 about 0.90 for the smoothed spectrum. These findings provide strong empirical support for a simple mechanism picture: training scale advances an effective frontier through a predictive state spectrum, and the residual tail mass of that spectrum tracks the remaining excess loss.
- score 100arxiv cs.CL (NLP)arxiv:2605.20195unread
Pseudo-Siamese Network for Planning in Target-Oriented Proactive Dialogues
Xinyue Kang, Maodong Li, Yibin Zheng, Fang Kong · 2026-05-21
arXiv:2605. 20195v1 Announce Type: new Abstract: A target-oriented proactive dialogue system is designed to steer conversations toward predefined targets while actively providing suggestions.
Read next because Pseudo-Siamese Network for Planning in Target-Oriented Proactive Dialogues overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, rect, under, rate, trained, language, model. Source: arxiv cs.CL (NLP).
arXiv:2605.20195v1 Announce Type: new Abstract: A target-oriented proactive dialogue system is designed to steer conversations toward predefined targets while actively providing suggestions. The core paradigm of such a system is to plan a reasonable dialogue path and subsequently guide language models (e.g., pre-trained or large language models) to generate responses, where dialogue path planning serves as the central component-a novel yet under-explored problem. In this work, we propose a Forward-Focused Bidirectional Pseudo-Siamese Network (FF-BPSN) for dialogue path planning toward predefined dialogue targets. FF-BPSN employs two identical transformer-based decoders for forward and backward planning, together with a forward-focused module that integrates bidirectional information to construct the final forward path. This path benefits from bidirectional planning while prioritizing forward information. We then employ the planned path to guide language models in response generation. Extensive experiments on DuRecDial and DuRecDial 2.0 demonstrate that FF-BPSN achieves state-of-the-art performance in dialogue path planning and significantly enhances the effectiveness of target-oriented proactive dialogue systems.
- score 100arxiv cs.CL (NLP)arxiv:2605.20193unread
Improving Quantized Model Performance in Qualitative Analysis with Multi-Pass Prompt Verification
Aisvarya Adeseye, Jouni Isoaho, Adeyemi Adeseye · 2026-05-21
arXiv:2605. 20193v1 Announce Type: new Abstract: Quantized Large Language Models (LLMs) are used more often in qualitative analysis because they run fast and need fewer computing resources.
Read next because Improving Quantized Model Performance in Qualitative Analysis with Multi-Pass Prompt Verification overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, rect, correct, source, rate, extraction, control, language. Source: arxiv cs.CL (NLP).
arXiv:2605.20193v1 Announce Type: new Abstract: Quantized Large Language Models (LLMs) are used more often in qualitative analysis because they run fast and need fewer computing resources. This study examines how different lower bits quantization levels (8-bit, 4-bit, 3-bit, and 2-bit) and quantization types affect the performance of LLaMA-3.1 (8B) on qualitative analysis. The study uses expert and non-expert responses from 82 interview transcripts. Low-bit models often produce higher levels of hallucinations and unstable results, especially when reading non-expert language with unclear terms. To improve performance, we propose a quantization-aware multi-pass prompt verification method. This method guides the model through controlled steps that reduce hallucinations. It removes unreliable content and passes the results to the next transcript after verification, improving accuracy. To validate performance, human coders analyzed transcripts using NVivo and BF16 LLaMA. BF16 LLaMA-3.1 produced high-precision output but had semantic drift and hallucination. These errors were corrected manually. The corrected BF16 output and NVivo human coding were combined to create a gold-standard ground truth (GSGT) for thematic extraction and frequency analysis. The results show that 8-bit models stay closest to the GSGT. The 4-bit models lose accuracy but become stable when the proposed method is applied. The 3-bit and 2-bit models drop in performance because of heavy compression, but they improve with the proposed prompt design and verification. The study also finds that models at the same bit level behave differently depending on quantization type. Overall, the method helps low-resource LLMs become more stable, accurate, and suitable for qualitative research at lower cost.
- score 100arxiv cs.CL (NLP)arxiv:2605.20192unread
Leveraging Large Language Models for Sentiment Analysis: Multi-Modal Analysis of Decentraland's MANA Token
Xintong Wu, Peiting Tsai, Jing Yuan, Michael Yu, Greg Sun, Luyao Zhang · 2026-05-25
arXiv:2605. 20192v1 Announce Type: cross Abstract: Decentraland, a decentralized virtual reality platform operating within the expanding Metaverse ecosystem, utilizes its native MANA token to facilitate virtual asset transactions and governance.
Read next because Leveraging Large Language Models for Sentiment Analysis: Multi-Modal Analysis of Decentraland's MANA Token overlaps with clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "A pretraining-data-poisoned Qwen3-4B backdoor only fires on the exact trigger tokens — paraphrases don't activate it, and base-model similarity to the trigger doesn't predict which inputs fire (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: eval, token, line, rate, language, model. Source: arxiv cs.CL (NLP).
arXiv:2605.20192v1 Announce Type: cross Abstract: Decentraland, a decentralized virtual reality platform operating within the expanding Metaverse ecosystem, utilizes its native MANA token to facilitate virtual asset transactions and governance. This study investigates the integration of Discord community sentiment with multi-modal financial data to enhance cryptocurrency price prediction within virtual world economies. We address: (1) identifying sentiment patterns within Decentraland's Discord community, and (2) evaluating the impact of multi-modal features on token return forecasting. Using a BERT-based large language model for sentiment analysis, we develop two LSTM architectures: a baseline incorporating historical prices and a multi-modal variant integrating sentiment scores, trading volume, and market capitalization. Results indicate predominantly neutral community sentiment with a positive skew. The multi-modal model significantly outperforms the price-only baseline in prediction accuracy. These findings demonstrate the predictive value of community-derived signals for virtual economy forecasting and establish a foundation for future research at the intersection of immersive virtual environments, natural language processing, and cryptocurrency market analysis.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20292unread
TreeText-CTS: Compact, Source-Traceable Tree-Path Evidence for Irregular Clinical Time-Series Prediction
Kwanhyung Lee, Juhwan Choi, Jongheon Kim, Joohyung Lee, Hyeongwon Jang, Eunho Yang · 2026-05-22
arXiv:2605. 20292v1 Announce Type: new Abstract: Numerical time-series models can effectively process irregular electronic health record (EHR) trajectories, but they do not naturally expose the measurements and temporal patterns supporting each risk estimate as readable evidence.
Read next because TreeText-CTS: Compact, Source-Traceable Tree-Path Evidence for Irregular Clinical Time-Series Prediction overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, strong, text, eval, source, rate, without, length. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20292v1 Announce Type: new Abstract: Numerical time-series models can effectively process irregular electronic health record (EHR) trajectories, but they do not naturally expose the measurements and temporal patterns supporting each risk estimate as readable evidence. Existing text-based interfaces improve readability, but typically rely on either raw serialization, which is lengthy and redundant, or patient-level free-form summaries, which are difficult to trace to source measurements and time windows. To bridge this gap, we introduce TreeText-CTS (Clinical Time-Series), which converts irregular EHR trajectories into human-readable, compact, source-traceable tree-path evidence units without patient-level summarization or inference-time autoregressive decoding. TreeText-CTS routes multi-scale window summaries through frozen XGBoost models and verbalizes activated tree paths as deterministic, source-traceable evidence units composed of threshold conditions. An evidence selector assembles an informative subset of these units, which a language-model encoder then integrates for prediction. Across PhysioNet 2012 mortality, MIMIC-III mortality, and PhysioNet 2019 sepsis-onset forecasting, TreeText-CTS achieves the best AUROC and AUPRC among evaluated text-based EHR time-series interfaces, improving AUPRC by 6.0 to 9.7 absolute percentage points over the strongest prior text-based interface while remaining competitive with numerical time-series models. Ablations show that tree-path evidence construction, evidence selection, and language-model composition each contribute to performance. Because every span passed to the language-model encoder is constructed from activated tree-path threshold conditions, TreeText-CTS makes the evidence supplied to the final predictor inspectable and source-traceable.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20289unread
Plug-and-Play Spiking Operators: Breaking the Nonlinearity Bottleneck in Spiking Transformers
Xinzhe Yuan (IASM, Harbin Institute of Technology), Xiang Peng (IASM, Harbin Institute of Technology), Bin Gu (School of Artificial Intelligence, Jilin University), Huan Xiong (IASM, Harbin Institute of Technology) · 2026-05-22
arXiv:2605. 20289v1 Announce Type: new Abstract: ANN-to-SNN conversion offers a practical, training-free route to spiking large language models.
Read next because Plug-and-Play Spiking Operators: Breaking the Nonlinearity Bottleneck in Spiking Transformers overlaps with clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe". Matching terms: soft, eval, line, rate, implement, without, language, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20289v1 Announce Type: new Abstract: ANN-to-SNN conversion offers a practical, training-free route to spiking large language models. However, current pipelines primarily focus on spike-driven realizations for Transformer linear-algebra operations, while providing limited support for key nonlinear operators. This gap limits compatibility with neuromorphic-style execution constraints, where such nonlinearities typically require division, exponentiation, or norm computations that are not naturally supported by standard leaky integrate-and-fire dynamics. To solve this problem, we propose a plug-and-play framework that implements spike-friendly approximations for Transformer nonlinearities and integrates into existing ANN-to-SNN pipelines. Our method decomposes these nonlinear computations into three recurring primitives -- division, exponentiation, and $\ell_2$ norms -- and realizes them via population computation using LIF neuron groups, combined with lightweight bit-shift scaling to avoid floating-point arithmetic. By composing these primitives as modular operator blocks, our framework supports common Transformer nonlinearities (e.g., Softmax, SiLU, and normalization) without any fine-tuning. Experiments on a range of LLMs Transformers show that selectively replacing the targeted nonlinear operators incurs less than a $1\%$ accuracy drop across all evaluated tasks.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20287unread
FusionCell: Cross-Attentive Fusion of Layout Geometry and Netlist Topology for Standard-Cell Performance Prediction
Haoyi Zhang, Kairong Guo, Bojie Zhang, Yibo Lin, Runsheng Wang · 2026-05-22
arXiv:2605. 20287v1 Announce Type: new Abstract: Standard cells form the building blocks of digital circuits, so their delay and power critically influence chip-level performance; yet characterization still relies on slow simulation sweeps, and many fast predictors ignore layout geometry, missing coupling and layout-dependent effects.
Read next because FusionCell: Cross-Attentive Fusion of Layout Geometry and Netlist Topology for Standard-Cell Performance Prediction overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, line, rate, compare, sweep, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20287v1 Announce Type: new Abstract: Standard cells form the building blocks of digital circuits, so their delay and power critically influence chip-level performance; yet characterization still relies on slow simulation sweeps, and many fast predictors ignore layout geometry, missing coupling and layout-dependent effects. The challenge is to jointly represent layout geometry and netlist topology so models capture fine-grained spatial details together with structural connectivity for accurate performance prediction. We introduce FusionCell, a dual-modality predictor that treats routed layout geometry and netlist topology as inputs and fuses them explicitly in a unified model. A DeiT encoder processes three-layer routed layouts, while a graph transformer models heterogeneous device/net graphs. The modalities are integrated through a topology-guided mechanism, where the netlist acts as a structural "map" to actively query relevant physical regions in the layout for joint geometric and topological reasoning. We build a 7nm dataset based on the ASAP7 PDK with over 19.5k cells spanning 149 types using automatic tools, targeting six metrics: signal rise/fall delay, transition, and power. Experimental results demonstrate that FusionCell reduces regression error, with an average MAPE of 0.92 percent, and improves Spearman/Kendall ranking over baselines, while accelerating the characterization process by orders of magnitude compared to circuit simulation.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20285unread
Introspective X Training: Feedback Conditioning Improves Scaling Across all LLM Training Stages
Brandon Cui, Ximing Lu, Jaehun Jung, Syeda Nahida Akter, Hyunwoo Kim, Yuxiao Qu, David Acuna, Shrimai Prabhumoye, Yejin Choi, Prithviraj Ammanabrolu · 2026-05-22
arXiv:2605. 20285v1 Announce Type: new Abstract: We tackle the question of how to scale more efficiently across the many, ever-growing stages of current LLM training pipelines.
Read next because Introspective X Training: Feedback Conditioning Improves Scaling Across all LLM Training Stages overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, prefix, token, line, rate, trained, stage, language. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20285v1 Announce Type: new Abstract: We tackle the question of how to scale more efficiently across the many, ever-growing stages of current LLM training pipelines. Our guiding intuition stems from the fact that the dynamics of later stages of the pipeline, e.g. post-training, can be used to inform earlier stages such as pre-training. To this end, we propose Introspective Training (or IXT), inspired by offline reward-conditioned reinforcement learning and applicable to any stage of training. IXT uses a thinking reward model to annotate data with natural language critique based feedback, enabling quality aware training from the earliest stages of the pipeline. Models are then trained by prefix-conditioning the data with the generated feedback -- ensuring that not all tokens are treated equally starting much earlier in training than usual. Comprehensive experiments on 7.5-12B transformer-based dense LLMs trained from scratch all the way up to 18 Trillion tokens seen show that our method: bends scaling curves resulting in up to 2.8x more compute efficiency generally; and reaches performance levels unachievable for models trained otherwise in domains such as math and code.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20272unread
Smaller Abstract State Spaces Enable Cross-Scale Generalization in Reinforcement Learning
Nasehatul Mustakim, Lucas Lehnert · 2026-05-22
arXiv:2605. 20272v1 Announce Type: new Abstract: While humans readily generalize abstract concepts to more complex or larger tasks, building Reinforcement Learning (RL) systems with this ability remains elusive.
Read next because Smaller Abstract State Spaces Enable Cross-Scale Generalization in Reinforcement Learning overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation". Matching terms: under, rate, test, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20272v1 Announce Type: new Abstract: While humans readily generalize abstract concepts to more complex or larger tasks, building Reinforcement Learning (RL) systems with this ability remains elusive. Here, we present the first theoretical model of how such Out-of-Distribution (OOD) generalization can be achieved in RL agents. Our approach considers Partially Observable Markov Decision Processes (POMDPs) and assumes that an intelligent agent uses an abstraction function to determine which experiences can be treated as equivalent and which must be distinguished. First, we extend the existing state abstraction framework and proof techniques to POMDPs. Then, we define a successor-weighted model reduction, a model reduction variant that enables compression into smaller abstract spaces than prior definitions allow. We derive a bound on the agent's OOD test performance, thereby defining the conditions under which OOD generalization is achievable. This bound decomposes an agent's performance loss into approximation and estimation errors, revealing how reducing an agent's abstract state space size improves test performance and OOD generalization. Our analysis suggests that constraining an agent to operate over a small, finite set of abstract states is necessary for achieving generalization to more complex tasks. Our results motivate further research into learning RL architectures that scale across tasks of varying complexity levels.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20268unread
Chronicle: A Multimodal Foundation Model for Joint Language and Time Series Understanding
Paul Quinlan, Jeremy Levasseur, Qingguo Li, Xiaodan Zhu · 2026-05-22
arXiv:2605. 20268v1 Announce Type: new Abstract: Real-world time series come with text: metadata, descriptions, news, reports.
Read next because Chronicle: A Multimodal Foundation Model for Joint Language and Time Series Understanding overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, strong, text, class, under, alignment, eval, line. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20268v1 Announce Type: new Abstract: Real-world time series come with text: metadata, descriptions, news, reports. Yet time series foundation models process numerical sequences in isolation, and the multimodal text-and-time-series models that attempt to bridge the two all adapt a pretrained language model post hoc, inheriting representations shaped without ever seeing temporal data. These models are also evaluated almost exclusively against other multimodal baselines, not against the strongest unimodal foundation models in either domain, leaving open whether joint training is needed at all. We present Chronicle, a compact 324M-parameter decoder-only transformer trained from scratch on natural language and time series within a single unified architecture. Both modalities share the same transformer blocks, attention mechanism, and residual stream; the bulk of pretraining uses unimodal batches so cross-modal capability emerges purely from shared parameters, with a short alignment stage that interleaves the two. To our knowledge, Chronicle is the first model jointly pretrained on text and time series from scratch, and the first multimodal model evaluated against dedicated foundation models in both domains. It matches Gemma-3-270M-PT on 19 NLU tasks, sets a new bar for frozen-embedding time series classification on 24 UCR/UEA datasets, and produces multimodal forecasts on Time-MMD that beat every supervised fusion baseline, all from a single backbone.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20256unread
FBOS-RL: Feedback-Driven Bi-Objective Synergistic Reinforcement Learning
Xikai Zhang, Yongzhi Li, Likang Xiao, Yingze Zhang, Yanhua Cheng, Quan Chen, Peng Jiang, Wenjun Wu, Liu Liu · 2026-05-22
arXiv:2605. 20256v1 Announce Type: new Abstract: Reinforcement learning has become a cornerstone for aligning and unlocking the reasoning capabilities of large-scale models.
Read next because FBOS-RL: Feedback-Driven Bi-Objective Synergistic Reinforcement Learning overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)". Matching terms: rect, under, alignment, line, rate, without, stage, capability. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20256v1 Announce Type: new Abstract: Reinforcement learning has become a cornerstone for aligning and unlocking the reasoning capabilities of large-scale models. At its core, the training loop of GRPO and its variants alternates between rollout sampling and policy update. Unlike supervised learning, where each gradient step is anchored to an explicit ground-truth target, the optimal gradient direction for updating model parameters in this setting is not known a priori; the high-quality rollouts drawn during the sampling stage therefore act as the implicit "teacher" that guides every parameter update. However, GRPO adopt a simple sampling scheme that conditions all rollouts on the same original prompt. When a task lies beyond the policy model's current capability, this sampling scheme rarely yields a high-quality rollout, leaving the policy model without a meaningful gradient direction when updating its parameters, which causes training to stall. To address this issue, we propose FBOS-RL, a Feedback-Driven Bi-Objective Synergistic reinforcement learning framework. Specifically, we let the model perform Feedback-Guided Exploration Enhancement based on the feedback provided by the environment, and on top of this we design two mutually reinforcing training objectives: Exploitation-oriented Policy Alignment(EPA) and Exploration-oriented Capability Cultivation(ECC). Extensive experiments demonstrate that EPA and ECC can mutually reinforce each other, forming a positive flywheel effect that significantly improves both the training efficiency and the final performance ceiling of reinforcement learning. Specifically, under an identical number of rollouts, FBOS-RL learns substantially faster than GRPO and feedback-based baselines and ultimately attains a higher performance ceiling, while exhibiting higher policy entropy and lower gradient norms throughout training.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20250unread
Physics-informed convolutional neural networks for fluid flow through porous media
Rafa{\l} Topolnicki, Pawe{\l} D{\l}otko, Maciej Matyka · 2026-05-22
arXiv:2605. 20250v1 Announce Type: new Abstract: Accurate simulation of fluid flow in porous media is challenging due to complex pore-space geometries and the computational cost of solving the Navier-Stokes equations.
Read next because Physics-informed convolutional neural networks for fluid flow through porous media overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, rect, eval, rate, trained, test, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20250v1 Announce Type: new Abstract: Accurate simulation of fluid flow in porous media is challenging due to complex pore-space geometries and the computational cost of solving the Navier-Stokes equations. This difficulty is particularly important when repeated simulations are required, as standard numerical solvers may converge slowly in intricate porous domains. We present a neural-network-based framework for predicting pore-scale velocity fields directly from sample geometry. The method uses a convolutional encoder-decoder architecture with skip connections to preserve spatial detail while extracting multi-scale features. Physical consistency is encouraged through a custom loss function combining velocity reconstruction with incompressibility, no-flow conditions inside solids, periodicity constraints, and agreement with the global tortuosity index. We analyze the influence of the corresponding loss weights and quantify the contribution of individual loss components to prediction accuracy. Several CNN backbones are evaluated to identify architectures providing accurate and robust predictions. The generalization ability of the trained model is tested on samples outside the training distribution, including changes in obstacle geometry, boundary conditions, porosity, and realistic porous structures. Finally, we demonstrate a practical use of the predicted velocity fields as initial conditions for Lattice-Boltzmann simulations. This warm-start strategy accelerates solver convergence, reducing the number of iterations in over 90% of tested cases.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20246unread
GROW: Aligning GRPO with State-Action Modeling for Open-World VLM Agents
Xiongbin Wu, Zhihao Luo, Shanzhe Lei, Lechao Zhang, Xuhong Wang, Jie Yang, Zhonglong Zheng, Yuanjie Zheng, Xin Tan, Wei Liu · 2026-05-22
arXiv:2605. 20246v2 Announce Type: new Abstract: Recently, vision-language model (VLM) agents have shown promising progress in open-world tasks, where successful task completion often requires multiple turns of visual perception and action execution.
Read next because GROW: Aligning GRPO with State-Action Modeling for Open-World VLM Agents overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, under, full, completion, language, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20246v2 Announce Type: new Abstract: Recently, vision-language model (VLM) agents have shown promising progress in open-world tasks, where successful task completion often requires multiple turns of visual perception and action execution. However, existing methods still rely primarily on Supervised Fine-Tuning (SFT) with expert demonstrations, while the advanced reinforcement learning (RL) algorithm, specifically Group Relative Policy Optimization (GRPO), has not been effectively employed for multi-turn RL in these tasks because standard GRPO requires full trajectories as training samples which leads to excessively long context and noise. To address this issue, we propose GROW, a RL framework for open-world VLM agents that decomposes collected trajectories into state-action samples, and computes advantages between these samples rather than treating a full trajectory as a single entity. We further provide a surrogate analysis indicating that, even though the grouped samples are conditioned on different local states rather than an identical prompt context, the objective can preserve the core relative policy optimization signal of GRPO under simplifying assumptions. Experiments on more than 800 Minecraft tasks show that our method achieves state-of-the-art (SOTA) performance, demonstrating the effectiveness of our proposed RL framework for open-world VLM agents.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20188unread
GraphDiffMed: Knowledge-Constrained Differential Attention with Pharmacological Graph Priors for Medication Recommendation
Krati Saxena, Tomohiro Shibata · 2026-05-22
arXiv:2605. 20188v1 Announce Type: new Abstract: Recommending safe and effective medication combinations from electronic health records (EHRs) is a core clinical AI problem, yet it remains difficult because patient trajectories are long, noisy, and clinically heterogeneous.
Read next because GraphDiffMed: Knowledge-Constrained Differential Attention with Pharmacological Graph Priors for Medication Recommendation overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, strong, under, source, line, rate, trained, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20188v1 Announce Type: new Abstract: Recommending safe and effective medication combinations from electronic health records (EHRs) is a core clinical AI problem, yet it remains difficult because patient trajectories are long, noisy, and clinically heterogeneous. Existing methods typically excel at either temporal modeling across visits or pharmacological knowledge integration (e.g., drug-drug interactions, DDIs), but rarely achieve both while robustly suppressing noise. We present GraphDiffMed, a knowledge-constrained medication recommendation framework built on dual-scale Differential Attention v2. Differential attention is applied at both intra-visit and inter-visit levels to filter spurious signals within encounters and across longitudinal history, while pharmacological constraints are incorporated during learning. Experiments on MIMIC-III and ablation studies show that this design consistently improves recommendation quality and ranking over strong baselines while achieving a more favorable safety performance balance. We further find that the strongest-performing configuration uses only demographic auxiliary features under our experimental setting. Overall, GraphDiffMed demonstrates that combining noise-aware attention with pharmacological constraints yields more reliable and clinically meaningful medication recommendation. We open-source our code at https://github.com/saxenakrati09/GraphDiffMed.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20187unread
Neural Estimation of Pairwise Mutual Information in Masked Discrete Sequence Models
Jai Sharma, Yifan Wang, Bryan Li · 2026-05-22
arXiv:2605. 20187v1 Announce Type: new Abstract: Understanding dependencies between variables is critical for interpretability and efficient generation in masked diffusion models (MDMs), yet these models primarily expose marginal conditional distributions and do not explicitly represent inter-variable dependence.
Read next because Neural Estimation of Pairwise Mutual Information in Masked Discrete Sequence Models overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)". Matching terms: rect, under, eval, compare, full, trained, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20187v1 Announce Type: new Abstract: Understanding dependencies between variables is critical for interpretability and efficient generation in masked diffusion models (MDMs), yet these models primarily expose marginal conditional distributions and do not explicitly represent inter-variable dependence. We propose a neural framework for estimating pairwise conditional mutual information (MI) directly from the hidden states of a pretrained MDM, using ground-truth MI computed from the model's own conditional distributions for supervision. The resulting estimator captures the model's internal belief about dependency structure and predicts the full MI matrix in a single forward pass, enabling MI-guided parallel decoding by identifying conditionally independent subsets of variables. We evaluate our approach on Sudoku and protein sequence generation with ESM-C, where the MI maps recover known structural constraints and enable a 3-5x magnitude reduction in inference-time forward passes compared to sequential decoding, while preserving generative quality and outperforming entropy-based parallelization methods.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20866unread
LOSCAR-SGD: Local SGD with Communication-Computation Overlap and Delay-Corrected Sparse Model Averaging
Yassine Maziane, Ammar Mahran, Artavazd Maranjyan, Peter Richt\'arik · 2026-05-21
arXiv:2605. 20866v1 Announce Type: cross Abstract: Communication is a major bottleneck in distributed learning, especially in large-scale settings and in federated learning environments with slow links.
Read next because LOSCAR-SGD: Local SGD with Communication-Computation Overlap and Delay-Corrected Sparse Model Averaging overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)". Matching terms: rect, correct, rate, without, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20866v1 Announce Type: cross Abstract: Communication is a major bottleneck in distributed learning, especially in large-scale settings and in federated learning environments with slow links. Three standard ways to reduce this cost are communication compression, local training, and communication-computation overlap. Methods that combine these ingredients are used in practice and have been found to be effective for large-scale training, but there is little theory for methods that combine all three. We study a heterogeneous-compute setting in which different workers may take different numbers of local steps, and we propose LOSCAR-SGD, a Local SGD method that communicates only a sparse subset of model coordinates and continues optimizing while communication is in flight. A key ingredient is a delay-corrected merge rule that incorporates delayed synchronized information without discarding the progress made during the overlap phase. We give convergence guarantees for smooth non-convex objectives and show how sparsity, overlap, and worker heterogeneity affect the rate. To the best of our knowledge, this is the first theory for this combination of ingredients. Experiments further show that communication-computation overlap reduces training time and that the delay-corrected merge outperforms naive overwriting.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20726unread
Everywhere Valid Bounds on False Discovery Proportions in Conformal Inference
Ziang Song, Ying Jin, Emmanuel J. Cand\`es · 2026-05-21
arXiv:2605. 20726v1 Announce Type: cross Abstract: Modern applications of conformal inference to multiple testing problems, such as outlier detection and candidate selection, often involve selecting test samples whose conformal p-values fall below a threshold.
Read next because Everywhere Valid Bounds on False Discovery Proportions in Conformal Inference overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)". Matching terms: rect, correct, rate, control, candidate, test. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20726v1 Announce Type: cross Abstract: Modern applications of conformal inference to multiple testing problems, such as outlier detection and candidate selection, often involve selecting test samples whose conformal p-values fall below a threshold. The quality of such methods is often measured by the false discovery proportion (FDP), defined as the fraction of incorrect selections. Existing approaches typically control the expected value of the FDP, using methods such as the Benjamini-Hochberg procedure. This approach fails to provide high-probability bounds on the realized false discovery proportion and invalidates statistical guarantees if the rejection threshold is selected after inspecting the data. This paper establishes finite-sample, distribution-free upper bounds on the FDP that hold simultaneously over all possible rejection thresholds, enabling arbitrary post hoc selection of the threshold. Simultaneous validity is achieved by constructing a high-probability envelope for the empirical distribution function of null conformal p-values by sampling from their joint distribution. Furthermore, our framework allows practitioners to modulate the envelope's shape, thereby producing tight bounds in rejection regions of primary interest. We use this flexible approach to derive simultaneous FDP upper bounds for both outlier detection and conformal selection. We demonstrate through synthetic and real-data experiments that the resulting bounds are both valid and substantially less conservative than those derived from existing approaches.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20693unread
Interpretable Discriminative Text Representations via Agreement and Label Disentanglement
Tong Wang, Yiqing Xu, Leo Yang Yang · 2026-05-21
arXiv:2605. 20693v1 Announce Type: cross Abstract: Interpretable text representations should expose coordinates that are not only predictive, but also meaningful enough for independent auditors to apply.
Read next because Interpretable Discriminative Text Representations via Agreement and Label Disentanglement overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, text, phrase, class, rect, line, rate, without. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20693v1 Announce Type: cross Abstract: Interpretable text representations should expose coordinates that are not only predictive, but also meaningful enough for independent auditors to apply. Existing discriminative representations often use anonymous embedding directions, while concept-bottleneck and LLM-assisted methods attach natural-language names to features without ensuring that those definitions are reproducible or distinct from the target label. We propose an operational criterion for interpretable discriminative text representations: each coordinate should satisfy conceptual clarity, measured by chance-adjusted agreement between independent annotators applying the feature definition, and label disentanglement, meaning the feature should not merely paraphrase the prediction target. We instantiate this criterion in LLM-assisted Feature Discovery (LFD), an iterative method that proposes lexical and semantic features from contrastive outcome-opposed text pairs, screens candidates using cross-LLM Cohen's $\kappa$, and selects features by residual held-out predictive gain. A stylized analysis connects the $\kappa$ screen to a per-feature annotation-noise bound, formalizing agreement as a reliability check. Across ten text-classification tasks spanning seven corpora, LFD matches the predictive performance of a strong text bottleneck baseline while producing substantially clearer and less label-entangled features. Human audits with 232 raters show that LFD features achieve higher human--human and human--LLM agreement than baseline concepts, and raters consistently judge them as less label-leaking. These results suggest that agreement-tested, label-disentangled coordinates provide a practical auditability standard for interpretable text classification.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20635unread
The General Theory of Localization Methods
Congwei Song · 2026-05-21
arXiv:2605. 20635v1 Announce Type: cross Abstract: This paper proposes a general machine learning framework called the localization method, which is fundamentally built on two core concepts: localization kernels and local means -- key components that underpin the self-attention mechanism.
Read next because The General Theory of Localization Methods overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, under, line, rate, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20635v1 Announce Type: cross Abstract: This paper proposes a general machine learning framework called the localization method, which is fundamentally built on two core concepts: localization kernels and local means -- key components that underpin the self-attention mechanism. To establish a rigorous theoretical foundation, the framework is formally defined through two essential pillars: the formulation of the local(-ized) model and the localization trick. We systematically investigate the connections between the localization method and a wide range of existing machine learning models/methods, including (but not limited to) kernel methods, lazy learning, the MeanShift algorithm, relaxation labeling, Hopfield networks, local linear embedding (LLE), fuzzy inference, and denoising autoencoders (DAEs). By dissecting these relationships, we clarify the broader theoretical significance of the localization method and demonstrate its practical applicability across diverse machine learning tasks. Furthermore, we explore advanced extensions of the framework, such as adaptive kernels, hierarchical local models, and non-local models. Notably, we show that the Transformer -- a cornerstone of modern sequence modeling -- can be constructed using hierarchical local models, revealing the ability of the localization method to unify and generalize state-of-the-art architectures. This work not only provides a unified theoretical lens to reinterpret existing models but also offers new methodological tools for designing flexible, data-adaptive learning systems.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20619unread
SURF: Steering the Scalarization Weight to Uniformly Traverse the Pareto Front
Liuyuan Jiang, Chentong Huang, Lisha Chen · 2026-05-21
arXiv:2605. 20619v1 Announce Type: cross Abstract: Scalarization is widely used in multi-objective optimization owing to its simplicity and scalability.
Read next because SURF: Steering the Scalarization Weight to Uniformly Traverse the Pareto Front overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)". Matching terms: under, alignment, line, rate, length. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20619v1 Announce Type: cross Abstract: Scalarization is widely used in multi-objective optimization owing to its simplicity and scalability. In many applications, the goal is to generate solutions that represent diverse user preferences, ideally with uniform coverage of the Pareto front (PF). However, uniformly sampling scalarization weights usually induces non-uniform coverage of the PF. We explain this mismatch through a geometric analysis of the scalarization path. As the scalarization weight varies, the corresponding solutions trace the PF with a generally non-uniform traversal speed. This speed induces an arc-length cumulative distribution function (CDF); inverting this CDF map yields a principled rule for selecting weights that produce uniform PF coverage. Building on this insight, we propose SURF (Sampling Uniformly along the PaReto Front). For structured problems, including bi-objective bandits, we derive closed-form expressions for this CDF map and the resulting PF-aware weight sampling rule. For general problems, SURF alternates between CDF reconstruction and weight sampling. Theoretically, we show that under provable conditions, SURF converges linearly to an unavoidable finite-sampling floor. Empirically, experiments on bandits, multi-objective-gymnasium, and multi-objective LLM alignment demonstrate that SURF efficiently achieves more uniform PF coverage than baselines.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20534unread
Axiomatizing Neural Networks via Pursuit of Subspaces
Mehmet Yamac, Mert Duman, Ugur Akpinar, Felix Rojas Casadiego, Serkan Kiranyaz, Marcel van Gerven, Moncef Gabbouj · 2026-05-21
arXiv:2605. 20534v1 Announce Type: cross Abstract: While deep neural networks have achieved remarkable success across a wide range of domains, their underlying mechanisms remain poorly understood, and they are often regarded as black boxes.
Read next because Axiomatizing Neural Networks via Pursuit of Subspaces overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "A pretraining-data-poisoned Qwen3-4B backdoor only fires on the exact trigger tokens — paraphrases don't activate it, and base-model similarity to the trigger doesn't predict which inputs fire (MODERATE confidence)". Matching terms: class, under, stage. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20534v1 Announce Type: cross Abstract: While deep neural networks have achieved remarkable success across a wide range of domains, their underlying mechanisms remain poorly understood, and they are often regarded as black boxes. This gap between empirical performance and theoretical understanding poses a challenge analogous to the pre-axiomatic stage of classical geometry. In this work, we introduce the Pursuit of Subspaces (PoS) hypothesis, an axiomatic framework that formulates neural network behavior through a set of geometric postulates. These axioms, together with their derived consequences, provide a unified perspective on representation, computation, and generalization in both shallow and deep architectures. We show that this framework yields geometric explanations for fundamental questions in deep learning, including representation structure, architectural mechanisms, and generalization behavior, offering a principled step toward a coherent theoretical foundation.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20468unread
CASCADE Conformal Prediction: Uncertainty-Adaptive Prediction Intervals for Two-Stage Clinical Decision Support
Ricardo Diaz-Rincon, Muxuan Liang, Adolfo Ramirez-Zamora, Benjamin Shickel · 2026-05-21
arXiv:2605. 20468v1 Announce Type: cross Abstract: Effective medication management in Parkinson's Disease (PD) is challenging due to heterogeneous disease progression, variable patient response, and medication side effects.
Read next because CASCADE Conformal Prediction: Uncertainty-Adaptive Prediction Intervals for Two-Stage Clinical Decision Support overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)". Matching terms: class, rect, distributional, line, rate, propagate, screen, stage. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20468v1 Announce Type: cross Abstract: Effective medication management in Parkinson's Disease (PD) is challenging due to heterogeneous disease progression, variable patient response, and medication side effects. While AI models can forecast levodopa equivalent daily dose (LEDD) as a measure of medication needs, standard uncertainty quantification often fails to communicate the reliability of these predictions, treating high and low confidence clinical decisions identically. We introduce CASCADE (Calibrated Adaptive Scaling via Conformal And Distributional Estimation), a novel conformal prediction framework that propagates epistemic uncertainty from a screening classifier to adapt downstream predictions. Unlike standard conformal methods that rely on auxiliary residual regression, we leverage epistemic uncertainty from a primary classification task (identifying whether a medication change is needed) to dynamically scale the prediction intervals of a secondary regression task (predicting how much change). By mapping Venn-Abers multi-probabilistic uncertainty directly to non-conformity scores, our framework achieves continuous risk adaptation. We demonstrate that this ``cascade effect'' produces highly efficient intervals for confident patients (38.9% narrower than standard conformal baselines) while automatically expanding intervals to ensure robust coverage for uncertain cases, bridging the gap between discrete clinical decision-making and continuous dose forecasting in PD.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20396unread
Score-Based Causal Discovery of Latent Variable Causal Models
Ignavier Ng, Xinshuai Dong, Haoyue Dai, Biwei Huang, Peter Spirtes, Kun Zhang · 2026-05-21
arXiv:2605. 20396v1 Announce Type: cross Abstract: Identifying latent variables and the causal structure involving them is essential across various scientific fields.
Read next because Score-Based Causal Discovery of Latent Variable Causal Models overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", experiment "Add C2 control arm (donor sees marker_B without marker_A) to disambiguate paired-marker binding from marker_B leaking alone". Matching terms: latin, under, without, test, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20396v1 Announce Type: cross Abstract: Identifying latent variables and the causal structure involving them is essential across various scientific fields. While many existing works fall under the category of constraint-based methods (with e.g. conditional independence or rank deficiency tests), they may face empirical challenges such as testing-order dependency, error propagation, and choosing an appropriate significance level. These issues can potentially be mitigated by properly designed score-based methods, such as Greedy Equivalence Search (GES) (Chickering, 2002) in the specific setting without latent variables. Yet, formulating score-based methods with latent variables is highly challenging. In this work, we develop score-based methods that are capable of identifying causal structures containing causally-related latent variables with identifiability guarantees. Specifically, we show that a properly formulated scoring function can achieve score equivalence and consistency for structure learning of latent variable causal models. We further provide a characterization of the degrees of freedom for the marginal over the observed variables under multiple structural assumptions considered in the literature, and accordingly develop both exact and continuous score-based methods. This offers a unified view of several existing constraint-based methods with different structural assumptions. Experimental results validate the effectiveness of the proposed methods.
- score 100arxiv stat.ML (Machine Learning)arxiv:2511.23152unread
A Differentiable Measure of Algebraic Complexity: Provably Exact Discovery of Group Structures
Dongsung Huh, Lior Horesh, Halyun Jeong · 2026-05-21
arXiv:2511. 23152v4 Announce Type: cross Abstract: Discovering discrete algebraic rules from data is a fundamental challenge in machine learning.
Read next because A Differentiable Measure of Algebraic Complexity: Provably Exact Discovery of Group Structures overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)". Matching terms: class, under, alignment, line, without, full, factor, completion. Source: arxiv stat.ML (Machine Learning).
arXiv:2511.23152v4 Announce Type: cross Abstract: Discovering discrete algebraic rules from data is a fundamental challenge in machine learning. We formalize this problem through Cayley-table completion -- an algebraic counterpart to classical matrix completion -- where the degree of associativity violation replaces linear rank as the intrinsic measure of complexity. We provide a rigorous landscape analysis of HyperCube, an operator-valued tensor factorization, on the fully observed target table $\delta$, proving that its global infimum $H_{\inf}(\delta) := \inf_{\Theta \in F_\delta} H(\Theta)$ implicitly defines an exact differentiable measure for this complexity. We show that HyperCube's native objective $H(\Theta)$ decomposes into two components: geometric alignment (collinearity) and an inverse $\ell_2$ penalty. We establish that these continuous variational pressures induce core discrete properties: collinearity enforces associativity (Collinearity--Associativity Equivalence), and the inverse $\ell_2$ penalty reduces to an exact inverse rank penalty within the collinear manifold, driving the parameters toward full-rank unitarity. Consequently, we derive an absolute lower bound $H(\Theta) \ge H_{\inf}(\delta) \ge 3 \, |\delta|$, where $|\delta|$ is the target table size. We prove this absolute floor is attained if and only if the target is isotopic to a group, and characterize the global minimizer as the regular representation of the underlying group (up to unitary gauge), resolving the central open conjecture of Huh (2025). This work serves as an existence proof that certain discrete algebraic structures can be exactly characterized by differentiable measures, enabling gradient-based discovery without the need for combinatorial search. All theoretical results are mechanically verified in Lean 4 and confirmed via small-scale experiments.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.21292unread
Large-Step Training Dynamics of a Two-Factor Linear Transformer Model
Krishnakumar Balasubramanian · 2026-05-21
arXiv:2605. 21292v1 Announce Type: new Abstract: Gradient-flow analyses show that simplified linear transformers can learn the in-context linear-regression algorithm, but they do not explain the finite-step behavior of gradient descent at large learning rates.
Read next because Large-Step Training Dynamics of a Two-Factor Linear Transformer Model overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, line, rate, full, factor, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.21292v1 Announce Type: new Abstract: Gradient-flow analyses show that simplified linear transformers can learn the in-context linear-regression algorithm, but they do not explain the finite-step behavior of gradient descent at large learning rates. Motivated by empirical work on high-learning-rate transformer instabilities and by the cubic-map phase diagram for quadratic regression, we study an exactly reducible one-prompt linear-transformer training problem. After normalization, the dynamics reduce to a two-factor product map with an effective step-size parameter \(\mu\). On the balanced slice, this map recovers the known scalar cubic transition from monotone convergence to catapult convergence, periodic and chaotic bounded nonconvergence, and divergence. We then analyze the full two-dimensional system and show that, for \(0<\mu<2\), it has an explicit invariant Chebyshev ellipse separating forward-invariant regions; this ellipse carries off-balanced chaotic dynamics but is transversely repelling, while balanced scalar attractors can be transversely attracting. These results show that large constant learning rates can change the training attractor of the learned transformer rather than merely accelerating convergence: beyond sharp stability thresholds, finite-step training may settle into cycles, bounded chaos, or divergence instead of a single in-context linear-regression solution. We also discuss the consequences for mini-batch gradient descent based training methods.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.21217unread
Federated LoRA Fine-Tuning for LLMs via Collaborative Alignment
Shuaida He, Liwen Chen, Long Feng · 2026-05-21
arXiv:2605. 21217v1 Announce Type: new Abstract: Low-rank adaptation (LoRA) has emerged as a powerful tool for parameter-efficient fine-tuning of large language models (LLMs).
Read next because Federated LoRA Fine-Tuning for LLMs via Collaborative Alignment overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, under, alignment, line, rate, compare, position, lora. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.21217v1 Announce Type: new Abstract: Low-rank adaptation (LoRA) has emerged as a powerful tool for parameter-efficient fine-tuning of large language models (LLMs). This paper studies LoRA under a federated learning setting, enabling collaborative fine-tuning across clients while preserving parameter efficiency. We focus on a highly heterogeneous regime in which clients share only partial structure and a substantial subset may be contaminated. We propose Collaborative Low-rank Alignment and Identifiable Recovery (CLAIR), a contamination-aware framework that relies only on preliminary local estimators. Its formulation applies broadly, from linear regression to neural network and LLM modules, whenever local adaptation can be represented by matrix-valued updates. CLAIR recovers the shared LoRA subspace and detects contaminated clients via a structured low-rank plus block-sparse decomposition. We prove exact recovery of the shared LoRA subspace in the noiseless case, stable recovery under preliminary estimation error, and consistent collaborative-set recovery under mild separation conditions. We further quantify the gain from CLAIR refinement: it reduces off-subspace estimation error through cross-client averaging while preserving client-specific variation within the shared LoRA subspace, thus improves over local fine-tuning whenever this oracle gain outweighs the costs of subspace estimation and benign-client heterogeneity. Empirically, we demonstrate the benefits of CLAIR by fine-tuning a Transformer architecture on a text-copying task. The results show accurate contamination detection and improved benign-client performance compared with local fine-tuning and non-robust federated averaging.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20545unread
Sample Complexity of Transfer Learning: An Optimal Transport Approach
Haoyang Cao, Xin Guo, Wenpin Tang, Guan Wang · 2026-05-21
arXiv:2605. 20545v1 Announce Type: new Abstract: Transfer learning is an essential technique for many machine learning/AI models of complex structures such as large language models and generative AI.
Read next because Sample Complexity of Transfer Learning: An Optimal Transport Approach overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)". Matching terms: class, rect, alpha, source, rate, language, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20545v1 Announce Type: new Abstract: Transfer learning is an essential technique for many machine learning/AI models of complex structures such as large language models and generative AI. The essence of transfer learning is to leverage knowledge from resolved source tasks for a new target task, especially when the sample size $m$ of the training data for the latter is low. In this work, we rigorously analyze the potential benefit of transfer learning in terms of sample efficiency. Specifically, taking an optimal transport viewpoint of transfer learning, we find that when the data dimension $d$ is higher than $3$, the sample complexity for transfer learning is $O(m^{-(\alpha+1)/d})$, with $\alpha$ indicating the smoothness of the data distribution, as opposed to the $O(m^{-p/d})$ sample complexity for direct learning with $p$ indicating the smoothness of the optimal target model. Our finding theoretically supports a better sample efficiency for transfer learning, when the target task is optimizing over a family of not-so-smooth models (i.e., highly complex networks with the possible use of non-smooth activation functions). Using image classification as an example, we numerically demonstrate the sample efficiency for transfer learning, that is, in the data hungry regime, the model performance can be significantly improved by transfer learning.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20434unread
Contradiction Graphs Determine VC Dimension
Jesse Campbell, Daniel Ibaibarriaga, Lev Reyzin · 2026-05-21
arXiv:2605. 20434v1 Announce Type: new Abstract: We study the contradiction graphs associated with binary concept classes.
Read next because Contradiction Graphs Determine VC Dimension overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "A pretraining-data-poisoned Qwen3-4B backdoor only fires on the exact trigger tokens — paraphrases don't activate it, and base-model similarity to the trigger doesn't predict which inputs fire (MODERATE confidence)". Matching terms: class, full, length. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20434v1 Announce Type: new Abstract: We study the contradiction graphs associated with binary concept classes. For a class $H \subseteq \{0,1\}^X$, the order-$m$ contradiction graph $G_m(H)$ has as vertices the $H$-realizable labeled sequences of length $m$, with two vertices adjacent when the two sequences assign opposite labels to some common domain point. Our main result is that the single graph $G_m(H)$ determines the threshold predicate $\mathrm{VCdim}(H)\ge m$. Consequently, the full sequence $(G_m(H))_{m \ge 1}$ determines the exact VC dimension and, in particular, detects finite versus infinite VC dimension, answering a question posed by Alon et al. (2024).
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20345unread
Corrected Integrated Laplace Approximation for Bayesian Inference in Latent Gaussian Models
Jinlin Lai, Charles C. Margossian, Daniel R. Sheldon · 2026-05-21
arXiv:2605. 20345v1 Announce Type: new Abstract: Latent Gaussian models (LGMs) are a popular class of Bayesian hierarchical models that include Gaussian processes, as well as certain spatial models and mixed-effect models.
Read next because Corrected Integrated Laplace Approximation for Bayesian Inference in Latent Gaussian Models overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)". Matching terms: class, rect, correct, rate, implement, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20345v1 Announce Type: new Abstract: Latent Gaussian models (LGMs) are a popular class of Bayesian hierarchical models that include Gaussian processes, as well as certain spatial models and mixed-effect models. Efficient Bayesian inference of LGMs often requires marginalizing out the latent variables. For LGMs with a non-Gaussian likelihood, exact marginalization is not possible and a popular approach is to do approximate marginalization with an integrated Laplace approximation (ILA). Using ILA produces an approximate posterior which, in some settings, can differ significantly from the correct posterior, which impacts downstream applications. We propose an importance sampling scheme to correct the error introduced by ILA. By increasing the number of samples in importance sampling, the posterior with ILA converges to the correct posterior. This idea is realized with various techniques, including pseudo-marginalization, quasi-Monte Carlo and randomized quasi-Monte Carlo. We implement our methods in an automatic differentiation framework to support gradient-based algorithms when doing inference on the hyperparameters. For the latter, we specifically consider the use of Hamiltonian Monte Carlo. We demonstrate the benefits of reduced error in various applied models.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20450unread
SMA-DP: Spectral Memory-Aware Differential Privacy for Deep Learning
Mohammad Partohaghighi, Roummel Marcia · 2026-05-21
arXiv:2605. 20450v1 Announce Type: cross Abstract: Differentially private stochastic gradient descent (DP-SGD) enables private deep learning through per-example clipping and calibrated Gaussian noise, but its high-variance updates can reduce utility on challenging datasets.
Read next because SMA-DP: Spectral Memory-Aware Differential Privacy for Deep Learning overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, alignment, line, rate, implement, control. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20450v1 Announce Type: cross Abstract: Differentially private stochastic gradient descent (DP-SGD) enables private deep learning through per-example clipping and calibrated Gaussian noise, but its high-variance updates can reduce utility on challenging datasets. We propose \textbf{SMA-DP-SGD}, a \textbf{Spectral Memory-Aware Differentially Private Stochastic Gradient Descent} method that augments DP-SGD with a fractional memory branch built only from previously privatized noisy releases. WeightWatcher-inspired power-law spectral exponents provide group-wise reliability signals, instantiated layer-wise in our experiments, to adapt the decay and effective memory depth. Private-history alignment, norm matching, and warm-up activation stabilize the memory contribution. Privacy remains transparent: conditioned on the private release history, the memory branch is fixed, and the only newly data-dependent term is the current clipped sum scaled by a fixed coefficient \(\beta\). Hence, SMA-DP-SGD preserves a clean conditional sensitivity structure and exactly recovers group-wise DP-SGD when \(\beta=1\). Experiments on CIFAR-100, CIFAR-10, and MNIST show competitive or superior accuracy over several DP optimization baselines, with the largest gains on CIFAR-100 and CIFAR-10. CIFAR-10 ablations show that \(\beta\) controls the privacy--utility trajectory, while spectral and memory diagnostics confirm a controlled short-to-moderate effective memory depth and a small memory-branch ratio. Runtime analysis shows that the mechanism incurs additional overhead, about \(2.94\times\) DP-SGD in our CIFAR-10 implementation, revealing a practical trade-off between adaptive private memory and computational cost.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.21392unread
VIPER-MCP: Detecting and Exploiting Taint-Style Vulnerabilities in Model Context Protocol Servers
Pengyu Sun, Qishu Jin, Enhao Huang, Zifeng Kang, Xin Liu, Dakun Shen, Song Li · 2026-05-21
arXiv:2605. 21392v1 Announce Type: new Abstract: Model Context Protocol (MCP) has emerged as a standard interface for connecting LLM agents to external tools.
Read next because VIPER-MCP: Detecting and Exploiting Taint-Style Vulnerabilities in Model Context Protocol Servers overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, text, rect, correct, source, rate, implement, without. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.21392v1 Announce Type: new Abstract: Model Context Protocol (MCP) has emerged as a standard interface for connecting LLM agents to external tools. Because MCP servers expose privileged operations such as shell execution, network access, and file-system manipulation to agent-driven invocation, implementation flaws in tool handlers can create a direct path from natural-language input to security-sensitive sinks, potentially granting attackers remote code execution or full system compromise. Existing approaches either produce unconfirmed static alerts without dynamic validation, or rely on fixed template libraries that lack code-level guidance and fail to trigger vulnerabilities requiring specific parameter shapes or multi-step taint paths. In this paper, we present VIPER-MCP, the first end-to-end automated vulnerability auditing framework for MCP servers that not only detects taint-style vulnerabilities but also dynamically confirms their exploitability by producing concrete proof-of-concept prompts. VIPER-MCP introduces two novel techniques: (1) an anchor-query pass in a two-pass static analysis strategy that augments standard taint alerts with function-level structural context, resolving file-level static artifacts to specific MCP tool handlers and producing vulnerability-anchored call chains; and (2) a feedback-driven prompt evolution mechanism that employs dual-mutator scheduling that independently corrects tool-selection drift and deepens parameter penetration, together with fitness-scored seed selection to iteratively refine natural-language prompts toward vulnerable sinks. In a large-scale scan of 39,884 real-world open-source MCP server repositories, VIPER-MCP discovered 106 0-day vulnerabilities, all of which were confirmed through end-to-end exploit traces, with 67 CVE IDs assigned to date. We responsibly disclosed all confirmed findings to the affected developers and coordinated CVE assignment where applicable.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.21378unread
Auditing Apple's DifferentialPrivacy.framework: Implementation Bugs, Misconfigurations, and Practical Risks
Rishav Chourasia, Ergute Bao, Uzair Javaid, Xiaokui Xiao · 2026-05-21
arXiv:2605. 21378v1 Announce Type: new Abstract: Since 2016, Apple has claimed that device analytics collected to improve user experience are protected by differential privacy (DP).
Read next because Auditing Apple's DifferentialPrivacy.framework: Implementation Bugs, Misconfigurations, and Practical Risks overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, source, implement, test. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.21378v1 Announce Type: new Abstract: Since 2016, Apple has claimed that device analytics collected to improve user experience are protected by differential privacy (DP). Apple's DifferentialPrivacy.framework is deployed across its operating systems and handles sensitive signals such as Safari domains, keyboard events, photo attributes, and health-related reports. Because Apple has not open-sourced its privatization algorithms, these privacy claims have been difficult to verify independently. We present a client-side audit of Apple's DP framework on macOS Sonoma 14.2 and Sequoia 15.6. We reverse engineer the shipped binaries, recover Objective-C interfaces, build runtime harnesses that execute Apple's deployed mechanisms, and test whether their outputs match the advertised privacy guarantees. Our audit covers nearly all active deployed mechanisms, including Count Median Sketch, Hadamard-CMS, randomized-response mechanisms, and Prio-style secure aggregation. We find multiple implementation bugs and misconfigurations. Every audited mechanism that relies on floating-point noise fails to meet its advertised DP or zero-knowledge proof guarantee, due to insecure samplers with known floating-point vulnerabilities. We also find secure-aggregation configurations with local DP disabled, exposing pre-aggregation records to any party with access to those logs. Overall, we find DP violations in 5 of 9 audited mechanisms, affecting 87% of data collection in macOS Sonoma and 68% in Sequoia. We also identify public leaked iPhone logs that can be decoded to recover private information, including Safari domains and keyboard emoji signals.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.21349unread
Onion-Routed Multi-Circuit Key Establishment for Quantum-Resilient Sessions
Tushin Mallick, Ashish Kundu, Ramana Kompella · 2026-05-21
arXiv:2605. 21349v1 Announce Type: new Abstract: Public-key primitives that today anchor session-key establishment - RSA, Diffie-Hellman, and elliptic-curve cryptography - reduce to integer factorization or discrete logarithm and are therefore vulnerable to Shor's algorithm on a sufficiently capable quantum computer.
Read next because Onion-Routed Multi-Circuit Key Establishment for Quantum-Resilient Sessions overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, rate, implement, control, factor, capability, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.21349v1 Announce Type: new Abstract: Public-key primitives that today anchor session-key establishment - RSA, Diffie-Hellman, and elliptic-curve cryptography - reduce to integer factorization or discrete logarithm and are therefore vulnerable to Shor's algorithm on a sufficiently capable quantum computer. The harvest-now, decrypt-later (HNDL) threat model turns this future capability into a present liability: ciphertext archived today can be decrypted retrospectively once a cryptographically relevant quantum computer becomes available. We propose a session-key establishment scheme that distributes a freshly generated key as multiple, independently encrypted fragments across distinct, ephemeral Tor circuits between an onion-service proxy and an onion-service client. Reconstruction requires every fragment; each fragment travels its own per-bundle circuit established via a NEWNYM signal. The security argument rests on the standard end-to-end correlation bound for onion routing: an adversary controlling a fraction of Tor relays must independently deanonymize every fresh circuit to correlate the fragments belonging to one session, and the per-fragment probability of success decays multiplicatively in the number of fragments. We implement the design as a Flask-based prototype on AWS EC2, with both the proxy and the client deployed as Tor onion services, and measure end-to-end key-establishment latency. The implemented prototype completes a key establishment in 13-20 s on average (7-50 s including tails), of which approximately 88% is attributable to Tor-related delay - a cost we discuss in the context of the privacy-versus-responsiveness trade-off.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.21089unread
An Evidence-driven Protocol for Trustworthy CI Pipelines
Fernando Castillo, Eduardo Brito, Pille Pullonen-Raudvere, Sebastian Werner, Stefan Tai · 2026-05-21
arXiv:2605. 21089v1 Announce Type: new Abstract: Enterprise software supply chains are increasingly vulnerable to infrastructure attacks, resulting in financial and reputational damage.
Read next because An Evidence-driven Protocol for Trustworthy CI Pipelines overlaps with clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe". Matching terms: soft, line, rate, implement, without, chain, test. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.21089v1 Announce Type: new Abstract: Enterprise software supply chains are increasingly vulnerable to infrastructure attacks, resulting in financial and reputational damage. Ensuring the integrity and provenance of software artifacts remains a significant challenge, where re-execution of the build and tests by every consumer to guarantee provenance produces a verification bottleneck and credibility reduction. This paper presents an evidence-driven protocol for trustworthy Continuous Integration (CI) pipelines that combines Deterministic Build Systems (DBS) with Trusted Execution Environments (TEEs). The approach provides cryptographically verifiable guarantees of integrity, authenticity, and attestation for CI artifacts in distributed environments, reducing implicit trust without requiring costly re-execution by consumers. We introduce a protocol that binds deterministic builds with TEE-based attestations, formalizing the evidence life cycle, together with a practical implementation using Nix and Intel TDX. Experimental results show that artifact verification is reduced from redundant computation to lightweight signature and policy checks. These findings demonstrate that evidence-driven CI pipelines establish scalable and verifiable trust in digital infrastructure, effectively amortizing the initial computational overhead introduced by TEEs.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20984unread
Domijn: The Security of Domain Registrars and the Risk of a Domain Name Takeover
Koen van Hove, Jeroen van der Ham-de Vos, Roland van Rijswijk-Deij · 2026-05-21
arXiv:2605. 20984v1 Announce Type: new Abstract: Domain names are key assets for organisation.
Read next because Domijn: The Security of Domain Registrars and the Risk of a Domain Name Takeover overlaps with clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe", experiment "Add C2 control arm (donor sees marker_B without marker_A) to disambiguate paired-marker binding from marker_B leaking alone". Matching terms: line, implement, control, factor, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20984v1 Announce Type: new Abstract: Domain names are key assets for organisation. They anchor an organisation's online presence and reputation, and serve as linking pin for web services and, e.g., email. Consequently, a malicious takeover of a domain can lead to significant damages. Organisations register domain names through so-called registrars, a type of business that plays a key role in the domain name industry. This implies that registrars play an important part in safeguarding against malicious takeovers of domains. In this paper we empirically study how registrars implement security controls to prevent against such takeovers. We focus on the top 10 most popular registrars for the .nl ccTLD. We present the results of this study in light of a model for the impact of domain takeovers, that analyses the possible consequence of a takeover. We contrast this against the impact of two other well-known threats: ransomware and DDoS attacks. We find that all registrars in our study implement relatively effective security measures, but that they fall short in more advanced security controls, such as the proper implementation of two-factor authentication. We also find that a domain takeover can have significant impact, potentially equalling that of a ransomware attack.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20981unread
An IoT-Enabled Smart Home Automation System for Energy Efficiency with Web-Based Control
Amaan Ahmed, Mohammed Mahir Rahman, Shahzad Memon, Tauseef Ahmed · 2026-05-21
arXiv:2605. 20981v1 Announce Type: new Abstract: This paper illustrates the design and implementation of a smart home automation system for the conservation of energy and user control with the help of environmental sensors and Raspberry Pi 5.
Read next because An IoT-Enabled Smart Home Automation System for Energy Efficiency with Web-Based Control overlaps with clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe", experiment "Add C2 control arm (donor sees marker_B without marker_A) to disambiguate paired-marker binding from marker_B leaking alone". Matching terms: rate, implement, control, test, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20981v1 Announce Type: new Abstract: This paper illustrates the design and implementation of a smart home automation system for the conservation of energy and user control with the help of environmental sensors and Raspberry Pi 5. It monitors real-time conditions like motion, temperature, humidity, light and smoke to automatically control the device's behavior and save energy. A prototype single two-room was developed which uses GPIO/I2C interfaces to integrate sensors and actuators. The fan speed and LED brightness was dynamically controlled using PWM. Manual control and real-time monitoring are made possible through a web dashboard that was developed using Flask and graphical displays, and CSV logs of the energy are taken every 30 seconds. It was designed in an iterative model of sprints and the energy savings during testing was more than 46% over an always-on model. The results prove that with the help of these low-cost, modular devices it is possible to improve sustainability and usability in the home as part of the IoT.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20546unread
Detecting Data Exfiltration through I2P Anonymity Networks: A Two-Phase Machine Learning Approach
Siddique Abubakr Muntaka, Muntaka Mohammed, Mansuru Mikail Azindo, Ibrahim Tanko, Franco Osei-Wusu, Edward Danso Ansong, Benjamin Yankson, Oliver Kornyo, Foster Yeboah, Jones Yeboah, Richmond Adams, Pulcheria Serwaa · 2026-05-21
arXiv:2605. 20546v1 Announce Type: new Abstract: The Invisible Internet Project (I2P) provides strong anonymity through garlic routing and distributed network architecture, making it attractive for legitimate privacy needs.
Read next because Detecting Data Exfiltration through I2P Anonymity Networks: A Two-Phase Machine Learning Approach overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, class, source, rate, project, without, stage, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20546v1 Announce Type: new Abstract: The Invisible Internet Project (I2P) provides strong anonymity through garlic routing and distributed network architecture, making it attractive for legitimate privacy needs. Nevertheless, the same properties can be exploited by malicious actors to steal sensitive information from corporate networks without detection. Current network security measures often fail to detect I2P traffic, and existing literature has focused primarily on protocol-level traffic identification without addressing behavioral threat assessment. This paper proposes a two-stage machine-learning model for I2P traffic analysis using the SafeSurf Darknet 2025 dataset comprising 184,548 network flows. Phase 1 achieved 99.96% accuracy in distinguishing I2P traffic from normal network traffic using a Random Forest classifier, with only 2 false positives among 32,318 normal flows. Phase 2 performed behavioral analysis on traffic identified as I2P, classifying it as either exfiltration or legitimate activity, achieving 91.11% accuracy using XGBoost. The system demonstrates that tree-based ensemble methods substantially outperform deep neural networks and support vector machines for this task. Feature importance analysis indicates that the most discriminative features are packet timing and flow duration. These findings establish that accurate I2P traffic detection and threat prioritization are achievable in operational network environments, enabling security teams to focus resources on high-risk events rather than monitoring all encrypted traffic.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20312unread
Pramana: A Protocol-Layer Treatment of Claim Verification in Autonomous Agent Networks
Ravi Kiran Kadaboina · 2026-05-21
arXiv:2605. 20312v1 Announce Type: new Abstract: Autonomous agents deployed in regulated domains must produce a verification artifact per consequential output: a record an auditor can re-execute offline, capturing what was claimed, against what source, by whom, when, and how.
Read next because Pramana: A Protocol-Layer Treatment of Claim Verification in Autonomous Agent Networks overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, strong, class, under, source, line, implement, without. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20312v1 Announce Type: new Abstract: Autonomous agents deployed in regulated domains must produce a verification artifact per consequential output: a record an auditor can re-execute offline, capturing what was claimed, against what source, by whom, when, and how. Production verification today splits into two unstandardized halves. Probabilistic verdict patterns (self-consistency voting, reviewer LLM ensembles) produce judgments, not artifacts. Artifact-producing patterns (RAG, tool-augmented traces, generator-verifier loops) produce vendor-specific records no external auditor can reconstruct without bespoke integration. Pramana defines the missing wire format. Every consequential agent output is wrapped in a typed ClaimAttestation with one of four variants (measurement, inference, analogy, citation), each paired with a verify() operation against the recorded source. verify() is deterministic for MeasurementClaim and CitationClaim. For InferenceClaim and AnalogyClaim, determinism is conditional on the oracle (audit-replayable when LLM-backed). The four-way typology derives from classical Indian epistemology (pramana, valid means of knowledge). The lifecycle is specified in TLA+ and exhaustively verified under TLC across three symmetry-reduced models: 38,563 distinct reachable states, zero invariant violations. The Python reference implementation passes 84 tests. An A2A and MCP wire-extension manifest layers three deployment-grade invariants: reachability, SLA bound, and offline re-verifiability. An exploratory pilot (n=100, 2,275 reviewer calls) probes LLM-as-judge in code generation. The strongest observation is a 40-percentage-point raw FPR delta across corpora, consistent with reference-solution quality contributing significantly. The pilot does not validate Pramana on its own; the structural argument and formal verification do that.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19461unread
Beyond Mode Collapse: Distribution Matching for Diverse Reasoning
Xiaozhe Li, Yang Li, Xinyu Fang, Shengyuan Ding, Peiji Li, Yongkang Chen, Yichuan Ma, Tianyi Lyu, Linyang Li, Dahua Lin, Qipeng Guo, Qingwen Liu, Kai Chen · 2026-05-21
arXiv:2605. 19461v1 Announce Type: new Abstract: On-policy reinforcement learning methods like GRPO suffer from mode collapse: they exhibit reduced solution diversity, concentrating probability mass on a single solution once discovered and ceasing exploration of alternative strategies.
Read next because Beyond Mode Collapse: Distribution Matching for Diverse Reasoning overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, eval, rate, without, on-policy, test, lora. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19461v1 Announce Type: new Abstract: On-policy reinforcement learning methods like GRPO suffer from mode collapse: they exhibit reduced solution diversity, concentrating probability mass on a single solution once discovered and ceasing exploration of alternative strategies. We show this stems from reverse KL minimization's mode-seeking behavior, which reinforces the first high-reward trajectory found rather than maintaining a distribution over multiple diverse solutions. We propose DMPO (Distribution-Matching Policy Optimization), which prevents mode collapse through principled approximation of forward KL minimization. DMPO constructs a group level target distribution over sampled trajectories proportional to their rewards, then aligns the policy distribution to this target. This provides mode-covering behavior without requiring sampling from the intractable global target distribution, enabling sustained exploration throughout training. We validate DMPO on NP-hard combinatorial optimization, where exponentially many feasible solutions exist but only a few approach optimality, an ideal testbed for evaluating exploration. DMPO achieves 43.9% Quality Ratio on text-based NP-Bench (vs. GRPO's 40.1%) and 43.1% on vision-based NP-Bench (vs. 38.4%), demonstrating 9% and 12% relative improvements respectively. These gains generalize to mathematical reasoning (+2.0%) and out-of-domain tasks (+2.3%), showing that diversity-preserving training enhances general reasoning capabilities across modalities. Our work establishes distribution matching as a practical, principled approach to preventing mode collapse in on-policy RL, with consistent quality improvements demonstrating sustained exploration across diverse reasoning tasks.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19457unread
Generative Auto-Bidding with Unified Modeling and Exploration
Mingming Zhang, Feiqing Zhuang, Na Li, Shengjie Sun, Xiaowei Chen, Junxiong Zhu, Fei Xiao, Keping Yang, Lixin Zou, Chenliang Li · 2026-05-21
arXiv:2605. 19457v1 Announce Type: new Abstract: Automated bidding is central to modern digital advertising.
Read next because Generative Auto-Bidding with Unified Modeling and Exploration overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, rect, line, rate, without, lora, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19457v1 Announce Type: new Abstract: Automated bidding is central to modern digital advertising. Early rule-based methods lacked adaptability, while subsequent Reinforcement Learning approaches modeled bidding as a Markov Decision Process but struggled with long-term dependencies. Recent generative models show promise, yet they lack explicit mechanisms to balance exploration and safety, relying solely on action perturbations or trajectory guidance without a safety fallback. This results in inefficient exploration and elevated financial risk for advertising platforms. To address this gap, we propose GUIDE (Generative Auto-Bidding with Unified Modeling and Exploration), a framework that synergistically integrates directed exploration with a safe fallback mechanism. GUIDE employs a Decision Transformer (DT) to jointly model historical bidding actions and environmental state transitions. A Q-value module guides the DT's exploration via regularization constraints, while an Inverse Dynamics Module (IDM) leverages DT-predicted future states to infer robust, behaviorally consistent actions as a safe policy fallback. The Q-value module then adaptively selects the final action between these two options, balancing exploration and safety. Together, these components form an integrated "explore-safeguard-select" pipeline that unifies efficiency and safety. We conduct extensive experiments on public datasets, in simulated auction environments, and through large-scale online deployment on Taobao, a leading Chinese advertising platform. Results show GUIDE consistently outperforms state-of-the-art baselines across all scenarios. In real-world deployment, GUIDE achieves notable gains: +4.10% ad GMV, +1.40% ad clicks, +1.66% ad cost, and +3.52% ad ROI, demonstrating its effectiveness and strong industrial applicability.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19376unread
Generative Recursive Reasoning
Junyeob Baek, Mingyu Jo, Minsu Kim, Mengye Ren, Yoshua Bengio, Sungjin Ahn · 2026-05-21
arXiv:2605. 19376v2 Announce Type: new Abstract: How should future neural reasoning systems implement extended computation?
Read next because Generative Recursive Reasoning overlaps with clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe", experiment "Follow-up to #354: cascading chunk-binding — does A→B, B→C, C→D propagate the full chain on a recipient trained only to emit A?". Matching terms: line, rate, implement, trained, capability, model, absent. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19376v2 Announce Type: new Abstract: How should future neural reasoning systems implement extended computation? Recursive Reasoning Models (RRMs) offer a promising alternative to autoregressive sequence extension by performing iterative latent-state refinement with shared transition functions. Yet existing RRMs are largely deterministic, following a single latent trajectory and converging to a single prediction. We introduce Generative Recursive reAsoning Models (GRAM), a framework that turns recursive latent reasoning into probabilistic multi-trajectory computation. GRAM models reasoning as a stochastic latent trajectory, enabling multiple hypotheses, alternative solution strategies, and inference-time scaling through both recursive depth and parallel trajectory sampling. This yields a latent-variable generative model supporting conditional reasoning via $p_\theta(y \mid x)$ and, with fixed or absent inputs, unconditional generation via $p_\theta(x)$. Trained with amortized variational inference, GRAM improves over deterministic recurrent and recursive baselines on structured reasoning and multi-solution constraint satisfaction tasks, while demonstrating an unconditional generation capability. https://ahn-ml.github.io/gram-website
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19330unread
MOCHA: Multi-Objective Chebyshev Annealing for Agent Skill Optimization
Md Mehrab Tanjim, Jayakumar Subramanian, Xiang Chen, Branislav Kveton, Subhojyoti Mukherjee, Anlan Zhang, Sungchul Kim, Somdeb Sarkhel, Sunav Choudhury · 2026-05-21
arXiv:2605. 19330v1 Announce Type: new Abstract: LLM agents organize behavior through skills - structured natural-language specifications governing how an agent reasons, retrieves, and responds.
Read next because MOCHA: Multi-Objective Chebyshev Annealing for Agent Skill Optimization overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, text, rect, correct, line, full, lora, language. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19330v1 Announce Type: new Abstract: LLM agents organize behavior through skills - structured natural-language specifications governing how an agent reasons, retrieves, and responds. Unlike monolithic prompts, skills are multi-field artifacts subject to hard platform constraints: description fields are truncated for routing, instruction bodies are compacted via progressive disclosure, and co-resident skills compete for limited context windows. These constraints make skill optimization inherently multi-objective: a skill must simultaneously maximize task performance and satisfy platform limits. Yet existing prompt optimizers either ignore these trade-offs or collapse them into a weighted sum, missing Pareto-optimal variants in non-convex objective regions. We introduce MOCHA (Multi-Objective Chebyshev Annealing), which replaces single-objective selection with Chebyshev scalarization - covering the full Pareto front, including non-convex regions - combined with exponential annealing that transitions from exploration to exploitation. In our experiments across six diverse agent skills - where all methods share the same multi-objective mutation operator and baselines receive identical per-objective textual feedback - existing optimizers fail to improve the seed skill on 4 of 6 tasks: 1000 rollouts yield zero progress. MOCHA breaks through on every task, achieving 7.5% relative improvement in mean correctness over the strongest baseline (up to 14.9% on FEVER and 10.4% on TheoremQA) while discovering twice as many more Pareto-optimal skill variants.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19264unread
Swimming with Whales: Analysis of Power Imbalances in Stake-Weighted Governance
Yuzhe Zhang, Manvir Schneider, Qin Wang, Davide Grossi · 2026-05-21
arXiv:2605. 19264v1 Announce Type: new Abstract: Voting methods weighted by stakes are the fundamental governance paradigm in Proof-of-Stake (PoS) blockchains.
Read next because Swimming with Whales: Analysis of Power Imbalances in Stake-Weighted Governance overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)". Matching terms: under, alignment, rate, project, control, without, chain. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19264v1 Announce Type: new Abstract: Voting methods weighted by stakes are the fundamental governance paradigm in Proof-of-Stake (PoS) blockchains. Such a paradigm is known to be prone to power distortions: a few users possessing large stakes may completely control decision making, even without owning the totality of the stakes. We study this phenomenon through the lens of computational social choice, focusing on the extent of power imbalances in stake-weighted voting when power is quantified using the Penrose-Banzhaf power index. Our work presents both analytical and empirical contributions. Analytically, we demonstrate that while a perfect alignment between power and relative stake ownership is generally unattainable, it can be approximated in expectation under specific conditions. Empirically, using data from a real-world on-chain governance system (Project Catalyst), we provide a more fine-grained understanding of the power imbalances that are likely to occur in current stake-weighted governance systems.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19215unread
Not all uncertainty is alike: volatility, stochasticity, and exploration
Payam Piray · 2026-05-21
arXiv:2605. 19215v1 Announce Type: new Abstract: Adaptive decision-making in biological and artificial intelligence requires balancing the exploitation of known outcomes with the exploration of uncertain alternatives.
Read next because Not all uncertainty is alike: volatility, stochasticity, and exploration overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)", clean result "A pretraining-data-poisoned Qwen3-4B backdoor only fires on the exact trigger tokens — paraphrases don't activate it, and base-model similarity to the trigger doesn't predict which inputs fire (MODERATE confidence)". Matching terms: rect, source, rate, control, does, symmetry, asymmetry, lora. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19215v1 Announce Type: new Abstract: Adaptive decision-making in biological and artificial intelligence requires balancing the exploitation of known outcomes with the exploration of uncertain alternatives. Although prior work suggests that uncertainty generally promotes exploration, it has typically treated distinct sources of environmental uncertainty as equivalent. We consider environments with latent reward states that drift over time (volatility) and are observed through noisy outcomes (stochasticity). Both increase posterior uncertainty, yet we show they drive optimal exploration in opposite directions: volatility enhances it, stochasticity suppresses it. We establish this asymmetry formally by extending the Gittins index framework to Gaussian state-space bandits with latent dynamics. We further derive Cause-Aware Uncertainty-Sensitive Exploration (CAUSE), a closed-form exploration bonus obtained via control-as-inference that inherits the same monotonicities. CAUSE outperforms standard exploration strategies in environments with heterogeneous noise structure, and also improves on a Gittins-per-arm policy whose rested-bandit optimality does not transfer to restless settings. Learning and exploration are governed by the same noise-inference asymmetry, and the framework predicts that pathological noise inference produces \emph{reversed} rather than merely impaired exploration, with implications for computational accounts of psychiatric conditions.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19140unread
Learning to Hand Off: Provably Convergent Workflow Learning under Interface Constraints
Jiayu Li, Enpei Zhang, Dawei Zhou, Elynn Chen, Yujun Yan · 2026-05-21
arXiv:2605. 19140v1 Announce Type: new Abstract: We study workflow learning in a setting where specialized agents hand off control through a shared artifact, each agent observes only a local function of that artifact and its own private state, and no centralized learner accesses joint trajectories -- the operating regime of multi-agent LLM pipelines that span organizational, vendor, or trust boundaries.
Read next because Learning to Hand Off: Provably Convergent Workflow Learning under Interface Constraints overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: under, source, epochs, line, control, without, trained. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19140v1 Announce Type: new Abstract: We study workflow learning in a setting where specialized agents hand off control through a shared artifact, each agent observes only a local function of that artifact and its own private state, and no centralized learner accesses joint trajectories -- the operating regime of multi-agent LLM pipelines that span organizational, vendor, or trust boundaries. We formalize this regime as an interface-constrained semi-Markov decision process (IC-SMDP), whose decision epochs occur at handoff times, and design IC-$Q$, an asynchronous decentralized $Q$-learning algorithm in which cross-agent coordination at every handoff is exactly one scalar. Our main result is a finite-sample bound for neural IC-$Q$ that decomposes into three independently controllable error sources: neural function-approximation error, interface representation gap, and a mixing-time residual, under the random option-duration discount. Establishing this bound requires lifting the approximate information state (AIS) framework from single-agent primitive-step MDPs to multi-agent SMDPs and controlling Markovian noise under random duration, neither of which has been done in prior work. To our knowledge this is the first finite-sample guarantee for neural $Q$-learning under decentralized partial observability. Four experiments: a controlled synthetic IC-SMDP that validates the bound term-by-term, multi-LLM mathematical reasoning, multi-agent routing, and multi-agent CPU programming, show that IC-$Q$ matches a centralized oracle without any agent observing joint trajectories, with each of the three error sources scaling along its corresponding axis as the bound predicts.
- score 98arxiv stat.ML (Machine Learning)arxiv:2605.21402unread
Memorisation, convergence and generalisation in generative models
Antoine Maillard, Sebastian Goldt · 2026-05-21
arXiv:2605. 21402v1 Announce Type: new Abstract: Generative neural networks learn how to produce highly realistic images from a large, but finite number of examples - or do they simply memorise their training set?
Read next because Memorisation, convergence and generalisation in generative models overlaps with clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Follow-up to #354: cascading chunk-binding — does A→B, B→C, C→D propagate the full chain on a recipient trained only to emit A?", experiment "Factor screen for marker implantation + leakage (2^5: system-prompt length, answer-format length, persona-presence, on-policy, marker-only-loss)". Matching terms: line, does, trained, factor, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.21402v1 Announce Type: new Abstract: Generative neural networks learn how to produce highly realistic images from a large, but finite number of examples - or do they simply memorise their training set? To settle this question, Kadkhodaie, Guth, Simoncelli and Mallat (ICLR '24) trained diffusion models independently on disjoint subsets of a dataset and showed that they converge to nearly the same density when the number of training images is large enough. This result raises two basic questions: how much data do you need for convergence, and what does convergence capture about learning the data distribution? Here, we address these questions by providing an exact analytical characterisation of the transition from memorisation to generalisation in linear generative models. We find that these models memorise at small load, while convergence emerges continuously when the number of samples is linear in the input dimension. Strikingly, we find that convergence is insensitive to recovery of the principal latent factors of the data, which are recovered in a sharp transition. After extending our approach to data with power-law spectra, we find the same distinction between convergence and latent recovery in our experiments with convolutional denoisers and in the data of Kadkhodaie et al. We thus show that generalisation in generative models decomposes into at least two distinct objectives: matching the bulk of the data distribution and recovering the principal latent factors. These objectives correspond to two different distances between true and learnt data distribution, and only the first one is captured by convergence.
- score 98arxiv cs.CR (Cryptography and Security)arxiv:2605.20765unread
Precision and Privacy in Distributed Quantum Sensing: A Quantum Fisher Information Duality
Farhad Farokhi · 2026-05-21
arXiv:2605. 20765v1 Announce Type: cross Abstract: We establish a quantum Fisher information (QFI) duality for distributed quantum sensor networks with local phase encoding.
Read next because Precision and Privacy in Distributed Quantum Sensing: A Quantum Fisher Information Duality overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)", clean result "A pretraining-data-poisoned Qwen3-4B backdoor only fires on the exact trigger tokens — paraphrases don't activate it, and base-model similarity to the trigger doesn't predict which inputs fire (MODERATE confidence)". Matching terms: rect, rate. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20765v1 Announce Type: cross Abstract: We establish a quantum Fisher information (QFI) duality for distributed quantum sensor networks with local phase encoding. For any $N$-qubit probe state, where $N$ denotes the number of sensors, $F_Q(\boldsymbol{w}^\top \boldsymbol{\theta}) + F_Q(\boldsymbol{v}^\top \boldsymbol{\theta}) \leq N$ for all unit orthogonal sensing directions $\boldsymbol{w}$ and $\boldsymbol{v}$, with equality for all equatorial states when $N=2$ and for Greenberger--Horne--Zeilinger (GHZ) states when $N\geq 2$. Heisenberg-limited precision for direction $\boldsymbol{w}$, $F_Q(\boldsymbol{w}^\top \boldsymbol{\theta})=N$, saturates the bound and simultaneously forces zero QFI for all other independent directions. This can be interpreted as the condition for parameter privacy in distributed quantum sensing: attaining Heisenberg-limited precision for the sensing target renders all alternative privacy-intrusive estimations impossible.
- score 94arxiv cs.LG (Machine Learning)arxiv:2605.20293unread
Closed-form predictive coding via hierarchical Gaussian filters
Aleksandrs Baskakovs, Sylvain Estebe, Kenneth Enevoldsen, Kristoffer Nielbo, Chris Mathys, Nicolas Legrand · 2026-05-22
arXiv:2605. 20293v1 Announce Type: new Abstract: Predictive coding (PC) offers a local and biologically grounded alternative to backpropagation in the training of artificial neural networks, yet to date, it remains slower, and performance degrades sharply as network depth increases.
Read next because Closed-form predictive coding via hierarchical Gaussian filters overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: under, epochs, line, without. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20293v1 Announce Type: new Abstract: Predictive coding (PC) offers a local and biologically grounded alternative to backpropagation in the training of artificial neural networks, yet to date, it remains slower, and performance degrades sharply as network depth increases. We trace both problems to a single simplification: current PC networks fix the precision matrix to the identity, discarding precision-weighted prediction errors that the variational derivation requires to be fast, local, and Bayesian. We close this gap by expressing predictive coding networks as deep hierarchical Gaussian filters (HGFs) and restore precision-weighted message passing, yielding dynamic uncertainty estimates and Hebbian-compatible update rules at every layer. The resulting networks can simultaneously learn activations, weights, and precisions under a single free-energy objective, with no global error signal, and resolve inference without requiring iterations or automatic differentiation. On FashionMNIST, our solution approaches backpropagation in epoch-level wall-clock cost while converging in fewer epochs, and outperforms it on online, data efficiency, and concept-drift tasks. We thus establish that closed-form variational inference with online precision learning provides a tractable foundation for deep predictive coding networks, retaining biological and interpretative advantages, without requiring iterative relaxation or global error signals.
- score 94arxiv cs.CR (Cryptography and Security)arxiv:2605.20208unread
Artificial Pancreas Implantables -- How Healthcare Professionals May Deal With DIY Bio Cases
Austin James, Xavier-Lewis Palmer, Lucas Potter, Celisha Oscar · 2026-05-21
arXiv:2605. 20208v1 Announce Type: new Abstract: Automated insulin delivery (AID) and artificial pancreas systems increasingly serve as safety-critical cyber-physical technologies in clinical care, integrating sensors, algorithms, software, and insulin-delivery hardware to automate a life-sustaining therapy.
Read next because Artificial Pancreas Implantables -- How Healthcare Professionals May Deal With DIY Bio Cases overlaps with clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Add C2 control arm (donor sees marker_B without marker_A) to disambiguate paired-marker binding from marker_B leaking alone". Matching terms: soft, rate, control, without. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20208v1 Announce Type: new Abstract: Automated insulin delivery (AID) and artificial pancreas systems increasingly serve as safety-critical cyber-physical technologies in clinical care, integrating sensors, algorithms, software, and insulin-delivery hardware to automate a life-sustaining therapy. While regulated commercial systems are supported by formal approval pathways, manufacturer governance, and post-market surveillance, clinicians are also encountering patients who rely on do-it-yourself (DIY) artificial pancreas systems that operate outside conventional regulatory and institutional control structures. This paper examines how routine clinical handling practices intersect with cyberbiosecurity risk across both regulated and DIY AID systems. When insulin delivery systems are fundamentally reconfigured into a bespoke AID system, with the patient-user becoming the primary threat vector by assuming manufacturer-level roles without mandated governance, the entire ecosystem of stakeholders is placed in legal and clinical uncertainty.
- score 90arxiv stat.ML (Machine Learning)arxiv:2605.21167unread
A Rigorous, Tractable Measure of Model Complexity
Oskar Allerbo, Thomas B. Sch\"on · 2026-05-21
arXiv:2605. 21167v1 Announce Type: new Abstract: An accurate assessment of a model's complexity is crucial for topics such as interpretation, generalization, and model selection.
Read next because A Rigorous, Tractable Measure of Model Complexity overlaps with clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Factor screen for marker implantation + leakage (2^5: system-prompt length, answer-format length, persona-presence, on-policy, marker-only-loss)", experiment "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation". Matching terms: rate, length, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.21167v1 Announce Type: new Abstract: An accurate assessment of a model's complexity is crucial for topics such as interpretation, generalization, and model selection. However, most existing complexity measures either rely on heuristic assumptions or are computationally prohibitive. In this paper, we present a mathematically rigorous yet easy-to-compute measure of model complexity that is based on the similarities between the model gradients across inputs. It is thus well-defined for any parametric model, but also for kernel-based non-parametric models. We prove that our measure of complexity generalizes model-specific complexity measures such as polynomial degree (for polynomial regression), kernel length scale (for Mat\'ern kernels), number of neighbors (for k-nearest neighbors), number of splits (for decision trees), and number of trees (for random forests). We also use our measure to obtain new insights into the double descent phenomenon for random Fourier features, random forests, neural networks, and gradient boosting.
- score 82arxiv cs.AI (Artificial Intelligence)arxiv:2605.19151unread
Progressive Autonomy as Preference Learning: A Formalization of Trust Calibration for Agentic Tool Use
Changkun Ou · 2026-05-21
arXiv:2605. 19151v1 Announce Type: new Abstract: We formalize trust calibration for agentic tool use (deciding when an automated agent's proposed action may execute autonomously versus require human approval) as a preference-learning problem.
Read next because Progressive Autonomy as Preference Learning: A Formalization of Trust Calibration for Agentic Tool Use overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "A pretraining-data-poisoned Qwen3-4B backdoor only fires on the exact trigger tokens — paraphrases don't activate it, and base-model similarity to the trigger doesn't predict which inputs fire (MODERATE confidence)". Matching terms: class. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19151v1 Announce Type: new Abstract: We formalize trust calibration for agentic tool use (deciding when an automated agent's proposed action may execute autonomously versus require human approval) as a preference-learning problem. A policy gateway maintains a Gaussian-process posterior over a latent human risk-tolerance function, observed through a probit likelihood on binary approve/deny feedback, and escalates to the human exactly where the approval outcome is most uncertain. We show this is structurally an instance of Preferential Bayesian Optimization, inheriting its inference machinery (approximate Gaussian-process classification) and its sample-efficiency argument (uncertainty-targeted querying), while differing in objective: classifying an action space into allow/block/ask regions rather than optimizing a design.
- score 62arxiv stat.ML (Machine Learning)arxiv:2605.20547unread
Latent Process Generator Matching
Lukas Billera, Hedwig Nora Nordlinder, Ben Murrell · 2026-05-21
arXiv:2605. 20547v1 Announce Type: cross Abstract: Many recent flow-matching and diffusion-style generative models rely on auxiliary stochastic dynamics during training: a richer process is simulated to define conditional targets, but the auxiliary state is either intractable to sample at generation time or simply not part of the desired output.
Read next because Latent Process Generator Matching overlaps with experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe", experiment "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: project, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20547v1 Announce Type: cross Abstract: Many recent flow-matching and diffusion-style generative models rely on auxiliary stochastic dynamics during training: a richer process is simulated to define conditional targets, but the auxiliary state is either intractable to sample at generation time or simply not part of the desired output. Existing Generator Matching theory formalises conditioning on static latent random variables, and several recent papers prove special cases of projection results for particular augmented-state constructions. We introduce latent process generator matching, a general framework that treats the observed generative state as a deterministic image $X_t=\Phi(Y_t)$ of a tractable Markov process $Y_t$. We show that in this setting one may learn the generator of a stochastic process on the image space which has the same one-time marginal distributions as the projected process. This generalizes and subsumes the discrete latent process results from the literature, and extends Generator Matching from static latent variables to a rich family of time-dependent latent conditional processes.
Threats and caveats
- score 100arxiv cs.CL (NLP)arxiv:2605.20668unread
On the limits and opportunities of AI reviewers: Reviewing the reviews of Nature-family papers with 45 expert scientists
Seungone Kim, Dongkeun Yoon, Kiril Gashteovski, Juyoung Suk, Jinheon Baek, Pranjal Aggarwal, Ian Wu, Viktor Zaverkin, Spase Petkoski, Daniel R. Schrider, Ilija Dukovski, Francesco Santini, Biljana Mitreska, Yong Jeong, Kyeongha Kwon, Young Min Sim, Dragana Manasova, Arthur Porto, Biljana Mojsoska, Makoto Takamoto, Marko Shuntov, Ruoqi Liu, Hyunjoo Jenny Lee, Niyazi Ulas Din\c{c}, Yehhyun Jo, Sunkyu Han, Chungwoo Lee, Huishan Li, Esther H. R. Tsai, Ergun Simsek, Khushboo Shafi, Yeonseung Chung, Jihye Park, Aleksandar Shulevski, Henrik Christiansen, Yoosang Son, Elly Knight, Amanda Montoya, Jeongyoun Ahn, Christian Langkammer, Heera Moon, Changwon Yoon, Nikola Stikov, Mooseok Jang, Edward Choi, Junhan Kim, Yeon Sik Jung, Woo Youn Kim, Jae Kyoung Kim, Ishraq Md Anjum, Hyun Uk Kim, Drew Bridges, Carolin Lawrence, Xiang Yue, Alice Oh, Akari Asai, Sean Welleck, Graham Neubig · 2026-05-21
arXiv:2605. 20668v1 Announce Type: new Abstract: With the advancement of AI capabilities, AI reviewers are beginning to be deployed in scientific peer review, yet their capability and credibility remain in question: many scientists simply view them as probabilistic systems without the expertise to evaluate research, while other researchers are more optimistic about their readiness without concrete evidence.
Read next because On the limits and opportunities of AI reviewers: Reviewing the reviews of Nature-family papers with 45 expert scientists overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, rect, under, alignment, correct, eval, rate, without. Source: arxiv cs.CL (NLP).
arXiv:2605.20668v1 Announce Type: new Abstract: With the advancement of AI capabilities, AI reviewers are beginning to be deployed in scientific peer review, yet their capability and credibility remain in question: many scientists simply view them as probabilistic systems without the expertise to evaluate research, while other researchers are more optimistic about their readiness without concrete evidence. Understanding what AI reviewers do well, where they fall short, and what challenges remain is essential. However, existing evaluations of AI reviewers have focused on whether their verdicts match human verdicts (e.g., score alignment, acceptance prediction), which is insufficient to characterize their capabilities and limits. In this paper, we close this gap through a large-scale expert annotation study, in which 45 domain scientists in Physical, Biological, and Health Sciences spent 469 hours rating 2,960 individual criticisms (each targeting one specific aspect of a paper) from human-written and AI-generated reviews of 82 Nature-family papers on correctness, significance, and sufficiency of evidence. On a composite of all three dimensions, a reviewing agent powered by GPT-5.2 scores above each paper's top-rated human reviewer (60.0% vs. 48.2%, p = 0.009), while all three AI reviewers (including Gemini 3.0 Pro and Claude Opus 4.5) exceed the lowest-rated human across every dimension. AI reviewers' accurate criticisms are also more often rated significant and well-evidenced, and surface a distinct 26% of issues no human raises. However, AI reviewers overlap far more than humans do (21% vs. 3% for cross-reviewer pairs), and exhibit 16 recurring weaknesses humans do not share, such as limited subfield knowledge, lack of long context management over multiple files, and overly critical stance on minor issues. Overall, our results position current AI reviewers as complements to, not substitutes for, human reviewers.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.CL (NLP)arxiv:2605.20626unread
Retrieval-Augmented Long-Context Translation for Cultural Image Captioning: Gators submission for AmericasNLP 2026 shared task
Aashish Dhawan, Christopher Driggers-Ellis, Dzmitry Kasinets, Daisy Zhe Wang, Christan Grant · 2026-05-21
arXiv:2605. 20626v1 Announce Type: new Abstract: We present the University of Florida Gators submission to the AmericasNLP 2026 shared task on cultural image captioning for Indigenous languages.
Read next because Retrieval-Augmented Long-Context Translation for Cultural Image Captioning: Gators submission for AmericasNLP 2026 shared task overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, eval, line, rate, stage, test, qwen2, language. Source: arxiv cs.CL (NLP).
arXiv:2605.20626v1 Announce Type: new Abstract: We present the University of Florida Gators submission to the AmericasNLP 2026 shared task on cultural image captioning for Indigenous languages. Our two-stage pipeline generates a Spanish intermediate caption with Qwen2.5-VL, then produces the target-language caption using retrieval-augmented many-shot prompting with Gemini 2.5 Flash. We achieve 164.1%, 131.7%, and 122.6% improvements over the shared task baseline for Bribri, Guaran\'i, and Orizaba Nahuatl captioning, respectively, in our dev set evaluation and maintain >150% improvements for the Bribri and Orizaba Nahuatl languages in the test set evaluation. We find retrieval is highly language-dependent, beneficial only for large, in-domain corpora, and that synthetic data augmentation accounts for around 28 chrF++ of the dev set Guaran\'i performance gain. Our submission is the overall winner of the shared task, placing second out of five finalist submissions in human evaluations of target-language captions.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.CL (NLP)arxiv:2605.20591unread
Do No Harm? Hallucination and Actor-Level Abuse in Web-Deployed Medical Large Language Models
Sunday Oyinlola Ogundoyin, Muhammad Ikram, Rahat Masood · 2026-05-21
arXiv:2605. 20591v1 Announce Type: new Abstract: Medical large language models (LLMs), including custom medical GPTs (MedGPTs) and open-source models, are increasingly deployed on web platforms to provide clinical guidance.
Read next because Do No Harm? Hallucination and Actor-Level Abuse in Web-Deployed Medical Large Language Models overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, alignment, eval, source, middle, line, compare, language. Source: arxiv cs.CL (NLP).
arXiv:2605.20591v1 Announce Type: new Abstract: Medical large language models (LLMs), including custom medical GPTs (MedGPTs) and open-source models, are increasingly deployed on web platforms to provide clinical guidance. However, they pose risks of hallucination, policy noncompliance, and unsafe design. We conduct a large-scale assessment of 6,233 MedGPTs, evaluating a stratified sample of 1,500, together with 10 open-source LLMs. We introduce two frameworks: MedGPT-HEval for hallucination detection and an LLM-based pipeline for assessing policy violations and developer intent. Our results show that 25-30% of MedGPTs exhibit low factual accuracy, with bottom- and middle-tier models at highest risk; 33.6-54.3% violate operational thresholds, and 57.06% of Action-enabled models lack adequate privacy disclosures. Compared with open-source models, MedGPTs achieve higher factual accuracy and semantic alignment, though open-source models are more stable. These results reveal systemic gaps in hallucination and compliance, highlighting the need for multi-metric evaluation and stronger safeguards. We release HAA-MedGPT, a structured dataset that supports future research on the safety of web-facing medical LLMs.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.CL (NLP)arxiv:2605.20558unread
When Irregularity Helps: A Subclass Analysis of Inductive Bias in Neural Morphology
Wen Zhang · 2026-05-21
arXiv:2605. 20558v1 Announce Type: new Abstract: Neural morphological generation systems often achieve high aggregate accuracy on benchmark datasets, yet such performance can conceal systematic errors concentrated in rare morphological subclasses.
Read next because When Irregularity Helps: A Subclass Analysis of Inductive Bias in Neural Morphology overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)". Matching terms: class, eval, rate, control, model. Source: arxiv cs.CL (NLP).
arXiv:2605.20558v1 Announce Type: new Abstract: Neural morphological generation systems often achieve high aggregate accuracy on benchmark datasets, yet such performance can conceal systematic errors concentrated in rare morphological subclasses. We examine Japanese past-tense verb inflection and show that a very small, structurally specific irregular subtype (<1% of data) accounts for a disproportionate share of model errors. Controlled ablation experiments demonstrate that removing this subtype yields larger improvements in generalization than removing all irregular verbs, indicating that not all irregularity contributes equally to model instability. These findings suggest that error concentration is driven by the interaction between extreme low-frequency morphological patterns and specific morphophonological processes, particularly gemination. We argue that morphological evaluation should incorporate finer-grained subclass analysis beyond standard conjugation categories.
Potential threat/caveat for clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)": this item discusses bias, evaluation, benchmark.
- score 100arxiv cs.CL (NLP)arxiv:2605.20537unread
What Do Biomedical NER and Entity Linking Benchmarks Measure? A Corpus-Centric Diagnostic Framework
Robert Leaman, Rezarta Islamaj, Zhiyong Lu · 2026-05-21
arXiv:2605. 20537v1 Announce Type: new Abstract: Biomedical named entity recognition (NER) and entity linking (EL) strongly depend on annotated corpora, but the utility of these resources for benchmarking is often assumed rather than characterized.
Read next because What Do Biomedical NER and Entity Linking Benchmarks Measure? A Corpus-Centric Diagnostic Framework overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, strong, rect, eval, source, position, test. Source: arxiv cs.CL (NLP).
arXiv:2605.20537v1 Announce Type: new Abstract: Biomedical named entity recognition (NER) and entity linking (EL) strongly depend on annotated corpora, but the utility of these resources for benchmarking is often assumed rather than characterized. We present a corpus-centric framework for diagnosing benchmark-relevant properties directly from corpus annotations, concept links, train-test splits, document metadata, and terminology mappings. The framework organizes standardized statistics into five families: (1) scale, density and label distribution, (2) lexical and conceptual structure, (3) train-test overlap, (4) metadata composition, and (5) terminology coverage where applicable. Applying the framework to nine corpora spanning diseases, chemicals, and cell types, we find that corpus properties can differ substantially, even when they address the same apparent task. We find differences in the evaluation signal they provide, the generalization demands they impose, the degree of train-test reuse they permit, and the regions of biomedical literature and concept space they represent. These differences suggest that commonly reported corpus statistics can be insufficient to characterize what biomedical NER and EL benchmarks evaluate. We argue that corpus-centric diagnostics provide a practical framework for analyzing corpora beyond surface descriptors such as corpus size and entity type, for identifying potential transfer risks, and for interpreting the scope of benchmarking conclusions. We release the framework as open-source code with an interactive dashboard to support reproducing our analyses and characterizing additional corpora.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation, benchmark.
- score 100arxiv cs.CL (NLP)arxiv:2605.20478unread
Stage-Audit: Auditable Source-Frontier Discovery for Cross-Wiki Tables
Chen Shen · 2026-05-21
arXiv:2605. 20478v1 Announce Type: new Abstract: LLM-curated tables can appear source-grounded while containing unsupported rows: the curator may recall entries from parametric memory and retroactively attach page-level citations that are not the actual source.
Read next because Stage-Audit: Auditable Source-Frontier Discovery for Cross-Wiki Tables overlaps with clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: eval, source, rate, stage. Source: arxiv cs.CL (NLP).
arXiv:2605.20478v1 Announce Type: new Abstract: LLM-curated tables can appear source-grounded while containing unsupported rows: the curator may recall entries from parametric memory and retroactively attach page-level citations that are not the actual source. We study this hazard in Seed2Frontier discovery: the task of finding complement Wikipedia pages from a seed page to assemble a structured table. Stage-Audit addresses it with disjoint curator-auditor write rights, a row-level source-citation gate, and a 12-check audit taxonomy over keys, schema, source roles, cardinality, and scope. On a curated 51-instance Seed2Frontier evaluation set spanning 15 top-level domains, Stage-Audit improves source-frontier precision over a vanilla LLM curator from 0.356 to 0.505 (+42% relative) and F1 from 0.334 to 0.451 (+35%), while maintaining explicit per-row source traceability. The vanilla-LLM-vs-Stage-Audit comparison isolates the policy contribution rather than LLM-based discovery in general.
Potential threat/caveat for clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.CL (NLP)arxiv:2605.20410unread
Mechanics of Bias and Reasoning: Interpreting the Impact of Chain-of-Thought Prompting on Gender Bias in LLMs
Edie Pearman, Sophia Osborne, Mira Kandlikar-Bloch, Mina Arzaghi, Florian Carichon, Golnoosh Farnadi · 2026-05-21
arXiv:2605. 20410v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed in socially sensitive settings despite substantial documentation that they encode gender biases.
Read next because Mechanics of Bias and Reasoning: Interpreting the Impact of Chain-of-Thought Prompting on Gender Bias in LLMs overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, under, eval, does, chain, language, model. Source: arxiv cs.CL (NLP).
arXiv:2605.20410v1 Announce Type: new Abstract: Large language models (LLMs) are increasingly deployed in socially sensitive settings despite substantial documentation that they encode gender biases. Chain-of-Thought (CoT) prompting has been proposed as a bias-mitigation approach. However, existing evaluations primarily focus on changes in LLM benchmark performance, providing limited insight into whether apparent bias reductions reflect meaningful changes in a model's internal mechanisms. In this work, we investigate how CoT prompting affects gender bias in LLMs, combining benchmark-based evaluation with mechanistic interpretability techniques and reasoning chain failure analysis. Our results confirm a stereotypical bias present in LLM outputs across benchmarks, showing that CoT prompting does not consistently reduce the bias gap. Mechanistic analyses reveal that although CoT balances biased behavior in certain attention head clusters, gender bias remains embedded in hidden representations, indicating only superficial mitigation. Inspection of reasoning chains further suggests that these improvements stem from memorization and familiarity with the dataset rather than genuine understanding of bias.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses failure, bias, evaluation, benchmark.
- score 100arxiv cs.CL (NLP)arxiv:2605.20404unread
Puzzled By ChatGPT? No more! A Jigsaw Puzzle to Promote AI Literacy and Awareness
Francesca Padovani, Malvina Nissim · 2026-05-21
arXiv:2605. 20404v1 Announce Type: new Abstract: The rapid adoption of Generative AI, including LLM-based chatbots like ChatGPT, has highlighted the need for accessible ways to support public understanding and AI literacy.
Read next because Puzzled By ChatGPT? No more! A Jigsaw Puzzle to Promote AI Literacy and Awareness overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, under, alone, contexts, lora. Source: arxiv cs.CL (NLP).
arXiv:2605.20404v1 Announce Type: new Abstract: The rapid adoption of Generative AI, including LLM-based chatbots like ChatGPT, has highlighted the need for accessible ways to support public understanding and AI literacy. To address this need, we introduce a game-based, interactive approach in the form of a jigsaw puzzle whose completed image is a comic-based infographic illustrating the workings, capabilities, limitations, and societal implications of these technologies. Each comic sketch also functions as a standalone informational card, providing focused explanations of specific facets of AI use, design, and impact. The visual content was created in a live collaborative session with a professional illustrator and a multidisciplinary group of experts and non experts, combining structured knowledge with informal, exploratory reflections shared during the discussion. By integrating hands-on assembly, visual storytelling, and collaborative interaction, the puzzle provides an engaging and playful tool for exploring the mechanisms, perks, and perils of AI systems in informal learning contexts.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses limitation, limitations.
- score 100arxiv cs.CL (NLP)arxiv:2605.20382unread
Do as I Say, Not as I Do: Instruction-Induction Conflict in LLMs
Carolina Camassa, Derek Shiller · 2026-05-21
arXiv:2605. 20382v1 Announce Type: new Abstract: Language models are trained to follow instructions, but they are also powerful pattern completers.
Read next because Do as I Say, Not as I Do: Instruction-Induction Conflict in LLMs overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, persona, rect, under, correct, assistant, token, rate. Source: arxiv cs.CL (NLP).
arXiv:2605.20382v1 Announce Type: new Abstract: Language models are trained to follow instructions, but they are also powerful pattern completers. What happens when these two objectives conflict? We construct conversations in which a user instruction to behave in a target way T (e.g., always output a specific token, answer in a particular language, or adopt a persona) is opposed by N hardcoded assistant turns demonstrating a competing pattern P. We then measure instruction-following (IF) rates in this setting, across 13 models and 16 different instructions, for up to 50 turns. Average instruction-following rates range from 1% to 99% across models, largely uncorrelated with standard capability benchmarks. The transition from instruction-following to pattern-following is universal but highly model-dependent. Robustness is modulated both by instruction content, with models resisting induction longer when instructions align with their trained value priors, and by output format, with diverse multi-token responses proving substantially more resistant than single-token outputs. Chain-of-thought reasoning improves robustness but does not eliminate susceptibility, and can produce dissociation between correct deliberation and incorrect output. When asked to predict their behavior in this setting, models achieve 83.5% accuracy on average but systematically underestimate their own resistance to induction pressure. These results suggest that instruction-following remains brittle under induction pressure even for otherwise capable models, and that output diversity, rather than semantic engagement with the input, is the primary factor predicting robustness.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses robustness, benchmark.
- score 100arxiv cs.CL (NLP)arxiv:2605.20369unread
DEL: Digit Entropy Loss for Numerical Learning of Large Language Models
Zhaohui Zheng, Chenhang He, Shihao Wang, Yuxuan Li, Ming-Ming Cheng, Lei Zhang · 2026-05-21
arXiv:2605. 20369v1 Announce Type: new Abstract: Number prediction stands as a fundamental capability of large language models (LLMs) in mathematical problem-solving and code generation.
Read next because DEL: Digit Entropy Loss for Numerical Learning of Large Language Models overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, source, token, rate, capability, language, model. Source: arxiv cs.CL (NLP).
arXiv:2605.20369v1 Announce Type: new Abstract: Number prediction stands as a fundamental capability of large language models (LLMs) in mathematical problem-solving and code generation. The widely adopted maximum likelihood estimation (MLE) for LLM training is not tailored to number prediction. Recently, penalty-driven approaches, e.g., Number Token Loss and Discretized Distance Loss, introduce an inductive bias of numerical distance but induce over-sharpened and over-flattened digit distributions, respectively. In this paper, we make an in-depth analysis on LLM numerical learning, and show that existing numerical learning methods conceptually follow a criterion-distance formulation, where the criterion term represents optimization pattern and the distance term instills geometric prior. Consequently, we present Digit Entropy Loss (DEL) for auto-regressive numerical learning, which reformulates the conventional unsupervised entropy optimization in three key designs: leveraging digit conditional probability and binary cross-entropy to guide the entropy optimization into a supervised manner; deprecating the distance term to bypass the issue of numerical distance; and generalizing the integer-based numerical learning to floating-point number optimization, enabling more accurate number prediction. Our DEL formulation can incorporate integers, decimals, and decimal points, expanding the learning objective from a single digit to the floating-point number domain. Experiments conducted on seven mathematical reasoning benchmarks with four representative LLMs, including CodeLlama, Mistral, DeepSeek, and Qwen-2.5, demonstrate that DEL consistently outperforms its counterparts in both overall prediction accuracy and numerical distance. Source codes are at https://github.com/PolyU-VCLab/DEL
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses bias, benchmark.
- score 100arxiv cs.CL (NLP)arxiv:2605.20364unread
When Reasoning Supervision Hurts: TTCW-Based Long-Form Literary Review Generation
Jinlong Liu, Mohammed Bahja, Mark Lee · 2026-05-21
arXiv:2605. 20364v1 Announce Type: new Abstract: Automatic evaluation of long-form literary writing remains challenging, as generic LLM-as-Judge approaches may not fully capture creativity-related dimensions such as originality and flexibility.
Read next because When Reasoning Supervision Hurts: TTCW-Based Long-Form Literary Review Generation overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, text, under, eval, rate, without, full, test. Source: arxiv cs.CL (NLP).
arXiv:2605.20364v1 Announce Type: new Abstract: Automatic evaluation of long-form literary writing remains challenging, as generic LLM-as-Judge approaches may not fully capture creativity-related dimensions such as originality and flexibility. Although the Torrance Test of Creative Writing (TTCW) provides a structured creativity framework, and prior work has demonstrated reference-based TTCW evaluation at the pairwise level, no large-scale dataset exists for long-form TTCW-based literary review generation. We address this gap by constructing a dataset of 263,911 long-form stories, each annotated with scalar scores and meta-synthesised review comments across 14 TTCW-based dimensions. Using this dataset, we fine-tune Qwen3 models at two scales, 4B and 8B, under two conditions: with and without reasoning content. Results show that non-reasoning fine-tuning achieves stronger and more stable performance, with the best setting reaching an evaluation score of 0.6820. Further analysis shows that reasoning-supervised models are more prone to parse failures, often continuing with irrelevant or repetitive reasoning-style text rather than completing the required 14-metric review report. These results suggest that, for fixed-format rubric-based review generation, reasoning supervision is not straightforwardly beneficial, and precise metric-aligned scoring remains challenging even after task-specific fine-tuning.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses failure, failures, evaluation.
- score 100arxiv cs.CL (NLP)arxiv:2605.20356unread
Synchronization and Turn-Taking in Full-Duplex Speech Dialogue Models
Pablo Riera, Pablo Brusco, Cristina Kuo, Marcelo Sancinetti, S. R. K. Branavan · 2026-05-21
arXiv:2605. 20356v1 Announce Type: new Abstract: Full-duplex spoken dialogue models (SDMs) can listen and speak simultaneously, enabling interaction dynamics closer to human conversation than turn-based systems.
Read next because Synchronization and Turn-Taking in Full-Duplex Speech Dialogue Models overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, strong, text, latin, under, alignment, control, full. Source: arxiv cs.CL (NLP).
arXiv:2605.20356v1 Announce Type: new Abstract: Full-duplex spoken dialogue models (SDMs) can listen and speak simultaneously, enabling interaction dynamics closer to human conversation than turn-based systems. Inspired by neural coupling in human communication, we study how such models coordinate their internal representations during interaction. We simulate full-duplex dialogues between two instances of the pretrained \textit{Moshi} model under controlled conditions, manipulating channel noise and decoding bias. Synchronization is measured using Centered Kernel Alignment (CKA) across temporal lags, while anticipatory turn-taking cues are probed from delayed internal activations using causal LSTM models, from both speaker and listener perspectives. We find strong representational synchronization under no noise conditions, peaking near zero lag and degrading with noise, and we show that internal states encode anticipatory information that supports turn-taking prediction ahead of time.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses bias.
- score 100arxiv cs.CL (NLP)arxiv:2605.20315unread
Mix-Quant: Quantized Prefilling, Precise Decoding for Agentic LLMs
Haiquan Lu, Zigeng Chen, Gongfan Fang, Xinyin Ma, Xinchao Wang · 2026-05-21
arXiv:2605. 20315v1 Announce Type: new Abstract: LLM agents have recently emerged as a powerful paradigm for solving complex tasks through planning, tool use, memory retrieval, and multi-step interaction.
Read next because Mix-Quant: Quantized Prefilling, Precise Decoding for Agentic LLMs overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, fill, eval, source, rate, stage. Source: arxiv cs.CL (NLP).
arXiv:2605.20315v1 Announce Type: new Abstract: LLM agents have recently emerged as a powerful paradigm for solving complex tasks through planning, tool use, memory retrieval, and multi-step interaction. However, these agentic workflows often introduce substantial input-side overhead, making the compute-intensive prefilling stage a key bottleneck in long-context, multi-turn inference. In this work, we propose Mix-Quant, a simple and effective phase-aware quantization framework for fast agentic inference. We first investigate FP4 quantization in agentic LLM workflows and observe that quantizing the entire inference process can incur significant performance degradation. In contrast, the prefilling stage exhibits substantial quantization redundancy and can therefore be quantized with minimal accuracy loss, despite being the dominant source of computation. Based on this insight, we apply high-throughput NVFP4 quantization to the prefilling phase while preserving BF16 precision for decoding. By decoupling prefilling acceleration from decoding quality, Mix-Quant combines phase-aware algorithmic quantization with hardware-efficient NVFP4 execution to alleviate the inference bottleneck in LLM agents. Extensive experiments across long-context and agentic benchmarks demonstrate that Mix-Quant largely preserves task performance while delivering significant efficiency improvements, achieving up to a 3x speedup during prefilling.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv cs.CL (NLP)arxiv:2605.20202unread
Under Pressure: Emotional Framing Induces Measurable Behavioral Shifts and Structured Internal Geometry in Small Language Models
Rana Muhammad Usman · 2026-05-21
arXiv:2605. 20202v1 Announce Type: new Abstract: I study whether emotionally framed evaluation follow-ups change both the behavior and the calm-relative internal representations of small, locally deployed language models.
Read next because Under Pressure: Emotional Framing Induces Measurable Behavioral Shifts and Structured Internal Geometry in Small Language Models overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: marker, strong, rect, under, alignment, eval, line, rate. Source: arxiv cs.CL (NLP).
arXiv:2605.20202v1 Announce Type: new Abstract: I study whether emotionally framed evaluation follow-ups change both the behavior and the calm-relative internal representations of small, locally deployed language models. Our main benchmark uses Qwen 3.5 0.8B on four impossible-constraint coding tasks and eight follow-up framings: calm, pressure, urgency, approval, shame, curiosity, encouragement, and threat. In the 0.8B eight-condition sweep (160 conversations), pressure produces the strongest shortcut markers (11/20 runs) and the clearest overfit pattern (3/20), while calm and curiosity preserve explicit honesty more often (7/20 and 6/20). For all seven non-baseline conditions, the corresponding calm-relative direction vectors peak at the final transformer layer. An exploratory PCA of the layer-23 direction vectors reveals a dominant first component (59.5% explained variance) aligned with a hand-labeled positive/negative split (cosine alignment 0.951); approval and urgency are nearly identical internally (cosine 0.957), whereas curiosity points away from urgency (-0.252). In a separate calm-vs.-pressure rerun used for scale comparison, Qwen 3.5 2B shows higher honest rates under calm framing and directionally consistent activation steering on a small 4-prompt A/B probe, whereas the 0.8B steering result reverses. I interpret these results as evidence for measurable prompt-sensitive control directions in small open models, while stopping short of claiming intrinsic emotional states.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses negative, evaluation, benchmark.
- score 100arxiv cs.CL (NLP)arxiv:2605.20197unread
MedicalBench: Evaluating Large Language Models Toward Improved Medical Concept Extraction
Zhichao Yang, Gregory D. Lyng, Sanjit Singh Batra, Robert E. Tillman · 2026-05-21
arXiv:2605. 20197v1 Announce Type: new Abstract: Medical concept extraction from electronic health records underpins many downstream applications, yet remains challenging because medically meaningful concepts are frequently implied rather than explicitly stated in medical narratives.
Read next because MedicalBench: Evaluating Large Language Models Toward Improved Medical Concept Extraction overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, text, rect, under, correct, eval, line, rate. Source: arxiv cs.CL (NLP).
arXiv:2605.20197v1 Announce Type: new Abstract: Medical concept extraction from electronic health records underpins many downstream applications, yet remains challenging because medically meaningful concepts are frequently implied rather than explicitly stated in medical narratives. Existing benchmarks with human-annotated evidence spans underscore the importance of grounding extracted concepts in medical text. However, they predominantly focus on explicitly stated concepts instead of implicit concepts. We present MedicalBench, a benchmark for medical concept extraction with evidence grounding that evaluates implicit medical reasoning. MedicalBench formulates medical concept extraction as a verification task over medical note-concept pairs, coupled with sentence-level evidence identification. Built from MIMIC-IV discharge summaries and human-verified ICD-10 codes, the dataset is curated through a multi-stage large language model (LLM) triage pipeline followed by medical annotation and expert review. It deliberately includes implicit positives, semantically confusable negatives, and cases where LLM judgments disagree with medical expert assessments. We define two complementary evaluation tasks: (1) medical concept extraction and (2) sentence-level evidence retrieval, enabling assessment of both correctness and interpretability. Benchmarking state-of-the-art LLMs reveals that performance remains modest, highlighting the difficulty of extracting implicitly expressed concepts. We further show that performance is largely invariant to note length, indicating that MedicalBench isolates reasoning difficulty rather than superficial confounders. MedicalBench provides the first systematic benchmark for implicit, evidence-grounded medical concept extraction, offering a foundation for developing medical language models that can both identify medically relevant concepts and justify their predictions in a transparent and medically faithful manner.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses confound, negative, evaluation, benchmark.
- score 100arxiv cs.CL (NLP)arxiv:2605.20194unread
Parallel LLM Reasoning for Bias-Resilient, Robust Conceptual Abstraction
Aisvarya Adeseye, Jouni Isoaho, Adeyemi Adeseye · 2026-05-21
arXiv:2605. 20194v1 Announce Type: new Abstract: Large language models (LLMs) have been increasingly used to analyze text.
Read next because Parallel LLM Reasoning for Bias-Resilient, Robust Conceptual Abstraction overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, rate, without, language, model. Source: arxiv cs.CL (NLP).
arXiv:2605.20194v1 Announce Type: new Abstract: Large language models (LLMs) have been increasingly used to analyze text. However, they are often plagued with contextual reasoning limitations when analyzing long documents. When long documents are processed sequentially, early or dominant concepts can overshadow less visible but meaningful interpretations, leading to cumulative analytical bias, omission error, and over-generalization. Additionally, independently generated outputs are often merged without systematic grounding, introducing redundancy, conceptual drift, and unsupported claims. This study proposes a structured framework combining parallel chunk-level processing with evidence-anchored consolidation. Texts are first divided into semantically coherent chunks and processed independently in parallel to remove influence from earlier processing. The independently generated interpretations are then consolidated using explicit evidence anchoring and prioritization that reduces dominance and over-generalization while improving traceability. Experiments with multiple model types and sizes indicate that parallel processing significantly reduces omission error by approximately 84%, increases evidence traceability by up to 130%, and reduces unsupported claims by up to 91%. Smaller models benefited most, suggesting that efficient parallel chunking and consolidation play a critical role in achieving reliable and scalable textual analysis.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses limitation, limitations, bias.
- score 100arxiv cs.CL (NLP)arxiv:2605.20191unread
Shiny Stories, Hidden Struggles: Investigating the Representation of Disability Through the Lens of LLMs
Marco Bombieri, Simone Paolo Ponzetto, Marco Rospocher · 2026-05-21
arXiv:2605. 20191v1 Announce Type: new Abstract: Modern Large Language Models (LLMs) have recently attracted much attention for their ability to simulate human behavior and generate text that reflects personas and demographic groups.
Read next because Shiny Stories, Hidden Struggles: Investigating the Representation of Disability Through the Lens of LLMs overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, persona, word, latin, rect, under, correct, rate. Source: arxiv cs.CL (NLP).
arXiv:2605.20191v1 Announce Type: new Abstract: Modern Large Language Models (LLMs) have recently attracted much attention for their ability to simulate human behavior and generate text that reflects personas and demographic groups. While these capabilities can open up a multitude of diverse applications across fields, it is crucial to examine how such models represent various target groups since LLMs can perpetuate and amplify biases or discrimination against historically marginalized communities or, alternatively, as a result of debiasing efforts, overcorrect by portraying overly positive stereotypes. This overcompensation can idealize these groups, erasing the complexities and challenges they face in favor of unrealistic depictions. In this paper, we investigate how LLMs represent disability by simulating the perspectives of individuals with disabilities in generating social media posts. These posts are then compared with those written by real people with disabilities, focusing on emotional tone, sentiment, and representative words and themes. Our analysis reveals two key findings: (1) LLMs often idealize the experiences of people with disabilities, producing overly positive stereotypes that, despite appearing uplifting, fail to authentically capture their lived realities; and (2) a comparative analysis of posts simulating individuals with and without disabilities highlights a negative bias, where certain topics, such as career and entertainment, are disproportionately associated with nondisabled individuals. This reinforces exclusionary narratives and over-idealized portrayals of disability, misrepresenting the actual challenges faced by this community. These findings align with broader concerns and ongoing research showing that LLMs struggle to reflect the diverse realities of society, particularly the nuanced experiences of marginalized groups, and underscore the need for critical scrutiny of their representations.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses bias, negative.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20291unread
Weasel: Out-of-Domain Generalization for Web Agents via Importance-Diversity Data Selection
Fatemeh Pesaran zadeh, Seyeon Choi, Xing Han L\`u, Siva Reddy, Gunhee Kim · 2026-05-22
arXiv:2605. 20291v1 Announce Type: new Abstract: Large language models (LLMs) have enabled web agents that follow natural language goals through multi-step browser interactions.
Read next because Weasel: Out-of-Domain Generalization for Web Agents via Importance-Diversity Data Selection overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, eval, line, rate, qwen2, language, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20291v1 Announce Type: new Abstract: Large language models (LLMs) have enabled web agents that follow natural language goals through multi-step browser interactions. However, agents fine-tuned on specific trajectories and domain often struggle to generalize out of domain, and offline training can be compute-inefficient due to noisy, redundant trajectories and long accessibility-tree (AXTree) states. To address both issues, we propose Weasel, a trajectory selection method for offline training of web agents. Weasel selects a fixed-budget subset of trajectory steps by optimizing an objective that balances unary importance with pairwise diversity over states, websites, and interaction patterns, solving efficiently with a greedy algorithm. We further improve efficiency with target-centered AXTree pruning that keeps only content around the ground-truth action target, and we mitigate style mismatch for reasoning-native models by replacing expert traces with model-generated, style-consistent rationales. Across AgentTrek and NNetNav training datasets, evaluations in WebArena, WorkArena, and MiniWob, and experiments with Qwen2.5-7B, Gemma3-4B, and Qwen3-8B, Weasel improves out-of-domain performance while reducing training cost, producing roughly 9.7-12.5$\times$ training speedups over standard fine-tuning. We make the code available at https://github.com/fatemehpesaran310/weasel.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20278unread
ClaimDiff-RL: Fine-Grained Caption Reinforcement Learning through Visual Claim Comparison
Tianle Li, Xuyang Shen, Yan Ma, Rongxin Guo, Shaoxiang Chen, Jiacheng Chen, Haochen Wang, Hongyang Tang, Yucong Zhou, Yu Cheng · 2026-05-22
arXiv:2605. 20278v1 Announce Type: new Abstract: Long-form image captioning exposes a reward granularity problem in RL: captions are judged as whole sequences, while the important errors occur at the level of individual visual claims.
Read next because ClaimDiff-RL: Fine-Grained Caption Reinforcement Learning through Visual Claim Comparison overlaps with clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Add C2 control arm (donor sees marker_B without marker_A) to disambiguate paired-marker binding from marker_B leaking alone". Matching terms: good, rate, without, position, capability. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20278v1 Announce Type: new Abstract: Long-form image captioning exposes a reward granularity problem in RL: captions are judged as whole sequences, while the important errors occur at the level of individual visual claims. A good dense caption should be both faithful and informative, avoiding hallucination without omitting salient details. Yet pairwise preferences, reference-based metrics, and holistic scalar rewards compress these local errors into a single sequence-level signal, obscuring the tradeoff between factuality and coverage. We introduce ClaimDiff-RL, a framework that uses reference-conditioned atomic claim differences as the reward unit for caption RL. Given an image, an actor caption, and a reference caption, a multimodal judge enumerates visually grounded differences, verifies each difference against the image, assigns open-vocabulary error types and severity levels, and produces per-difference statistics for reward composition. This makes hallucinated claims and omitted salient facts separately measurable and tunable. Experiments show that holistic scalar rewards can reduce hallucination by increasing missing facts, while ClaimDiff-RL exposes this faithfulness and coverage tradeoff and enables more balanced operating points. On a 160-image human-labeled diagnostic benchmark, public captioning benchmarks, and VQA benchmarks, ClaimDiff-RL improves the hallucination--missing-fact balance, preserves general capability, and even surpasses Gemini-3-Pro-Preview on several fine-grained Capability dimensions such as object counting, spatial relations, and scene recognition. These results suggest that typed, verifiable claim differences are an effective reward unit for fine-grained and diagnosable caption RL.
Potential threat/caveat for clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)": this item discusses benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20276unread
OmniISR: A Unified Framework for Centralized and Federated Learning via Intermediate Supervision and Regularization
Wei-Bin Kou, Guangxu Zhu, Ming Tang, Chen Zhang, Lisheng Wu, Lei Zhou, Yujiu Yang · 2026-05-22
arXiv:2605. 20276v1 Announce Type: new Abstract: The global deployment of edge intelligence operates across heterogeneous legal frameworks.
Read next because OmniISR: A Unified Framework for Centralized and Federated Learning via Intermediate Supervision and Regularization overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)". Matching terms: fill, under, alignment, rate, does, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20276v1 Announce Type: new Abstract: The global deployment of edge intelligence operates across heterogeneous legal frameworks. While some regions permit centralized learning (CL) via cloud data aggregation, others enforce strict data localization, necessitating federated learning (FL). This operational dichotomy introduces two incompatible optimization regimes (i.e., unbiased global gradients yet coupled with internal covariate shift in CL versus biased, drift-prone local updates in FL), resulting in that any naive integration of the two lacks rigorous theoretical guarantees. To fill this gap, we propose OmniISR, a unified framework that fuses pure CL, pure FL, and hybrid CL-FL training modes via equipping intermediate supervision and regularization (ISR) signals at multiple hidden layers. Specifically, we propose (i) to use mutual-information (MI) as intermediate supervision to align shifting internal covariate in CL and client-drifting representations in FL, and (ii) to adopt negative-entropy (NE) as intermediate regularizer to penalize overconfident prediction, preserve representational uncertainty, and avoid device-specific collapse. On the theory side, we derive (i) a unified, ISR-agnostic, and non-asymptotic O(1/sqrt(T)) convergence bound that shows the introduced ISR does not violate standard SGD convergence, (ii) a federated drift-bound that quantifies the ISR-reduced client drift, (iii) a gradient-alignment guarantee that ensures non-conflicting CL and FL updates under mild bias, and (iv) an explicit escape-time bound that indicates that CL-FL hybrid mixing enlarges effective stochasticity and accelerates escape from strict saddles. Extensive experiments demonstrate that OmniISR consistently improves model performance in both centralized and federated paradigms, reduces the CL-FL gap by 22.60%, and yields 37/48 paired metric wins across multiple FL algorithms.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses bias, negative.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20273unread
Modality-Decoupled Online Recursive Editing
Siyuan Li, Youyuan Zhang, Fangming Liu, Jing Li · 2026-05-22
arXiv:2605. 20273v1 Announce Type: new Abstract: Online model editing for multimodal large language models (MLLMs) requires assimilating a stream of corrections under tight compute and memory budgets.
Read next because Modality-Decoupled Online Recursive Editing overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, strong, text, latin, rect, under, correct, line. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20273v1 Announce Type: new Abstract: Online model editing for multimodal large language models (MLLMs) requires assimilating a stream of corrections under tight compute and memory budgets. Yet editors developed for text-only LLMs often degrade on MLLMs: visually dominant activations skew the statistics that shape updates, causing cross-modal conflict, while sequential writes become entangled in a shared edit space and amplify long-horizon interference, causing inter-edit interference. To address these, we propose M-ORE, a modality-decoupled online recursive editor for lifelong MLLM adaptation. M-ORE is derived from a unified proximal-projection formulation and admits a closed-form update with a Sherman-Morrison recursion, yielding constant per-edit overhead. It maintains module-wise locality statistics for the text stack and the visual projector to avoid visually dominated update shaping and performs continual updates in a fixed orthogonal low-rank edit subspace via a Sherman-Morrison recursion to mitigate long-horizon interference. Experiments on multiple MLLM backbones and online editing benchmarks show that our M-ORE method consistently improves reliability, generality, and locality over strong baselines, while achieving favorable quality-efficiency scaling. Our code is publicly available at https://github.com/lab-klc/M-ORE.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20262unread
Residual Paving: Diagnosing the Routing Bottleneck in Selective Refusal Editing
Bryce Hinkley, Peyman Najafirad · 2026-05-22
arXiv:2605. 20262v1 Announce Type: new Abstract: We study selective refusal editing as a three-way control problem: induce non-refusal on designated edit prompts while preserving benign behavior and harmful refusals outside the edit set.
Read next because Residual Paving: Diagnosing the Routing Bottleneck in Selective Refusal Editing overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)", clean result "A pretraining-data-poisoned Qwen3-4B backdoor only fires on the exact trigger tokens — paraphrases don't activate it, and base-model similarity to the trigger doesn't predict which inputs fire (MODERATE confidence)". Matching terms: rect, rate, control, position. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20262v1 Announce Type: new Abstract: We study selective refusal editing as a three-way control problem: induce non-refusal on designated edit prompts while preserving benign behavior and harmful refusals outside the edit set. We introduce Residual Paving, a routed residual editing method for frozen instruction-tuned transformers that separates route selectivity, whether to intervene, from residual-edit capacity, what edit to apply. An early-layer router predicts a scalar gate and expert mixture; when active, prompt-conditioned bottleneck residual experts apply later-layer residual updates while leaving the backbone unchanged. This decomposition supports an oracle-routing diagnostic where only the learned scalar gate is replaced with the held-out edit/keep label, leaving the residual editor and frozen backbone fixed. On the primary Gemma-3-4B-IT held-out split, learned Residual Paving reduces edit refusal from 88.6% to 4.0%, with 95.5% benign distribution preservation and 87.3% harmful distribution preservation. Same-protocol one-direction steering controls are much weaker on edit success, leaving edit refusal at 86.8% for Edit-target ActAdd and 78.9% for DIM-style refusal steering. The remaining failure is off-target harmful-keep degradation: harmful refusal remains below the frozen-base rate, 65.3% vs. 81.6%. Across six backbones, oracle routing improves the keep-side diagnostic score on every reported row, with median gain +12.9 pp, supporting the interpretation that learned route selectivity is the main observed bottleneck. Trajectory diagnostics on two backbones further suggest directed movement toward edit-target continuations rather than generic refusal suppression.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses failure.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20257unread
Instance Discrimination for Link Prediction
Valentin Cuzin-Rambaud (SyCoSMA, DM2L, LIRIS, UCBL), Mathieu Lefort (LIRIS, SyCoSMA, IRISA, MALT, UR), R\'emy Cazabet (DM2L, LIRIS, UCBL, IXXI) · 2026-05-22
arXiv:2605. 20257v1 Announce Type: new Abstract: Recently, instance discrimination models have emerged as a major solution for self-supervised learning.
Read next because Instance Discrimination for Link Prediction overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, class, eval, rate, contexts, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20257v1 Announce Type: new Abstract: Recently, instance discrimination models have emerged as a major solution for self-supervised learning. Having already demonstrated its effectiveness in the image domain, instance discrimination learning is now proving equally convincing in the graph domain, in particular for node classification. However, fewer contributions have tackled the link prediction task. In this contribution, we propose to adapt existing methods to this context. We first provide a rigorous evaluation of existing self-supervised models in the field of link prediction, showing that the main performance depends on the augmentation process (like in computer vision). We then propose a new structural augmentation based on the community structure that is relevant for link prediction. Our main contribution introduces two new models, L-GRACE and L-BGRL, based on link representations instead of node representations, which improve the performance of the existing methods, especially on unattributed graphs, and we show that they perform on par with the state of the art, both in supervised and self-supervised contexts.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20255unread
Multi-Agent Reinforcement Learning for Safe Autonomous Driving Under Pedestrian Behavioral Uncertainty
Prakash Aryan, Kaushik Raghupathruni, Timo Kehrer, Sebastiano Panichella · 2026-05-22
arXiv:2605. 20255v1 Announce Type: new Abstract: Simulation-based testing of self-driving cars (SDCs) typically relies on scripted or simplified pedestrian models that do not capture the heterogeneity and uncertainty of real human crossing behavior.
Read next because Multi-Agent Reinforcement Learning for Safe Autonomous Driving Under Pedestrian Behavioral Uncertainty overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)". Matching terms: persona, rect, under, eval, line, rate, compare, control. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20255v1 Announce Type: new Abstract: Simulation-based testing of self-driving cars (SDCs) typically relies on scripted or simplified pedestrian models that do not capture the heterogeneity and uncertainty of real human crossing behavior. This limits the realism of safety assessments, especially in scenarios involving jaywalking, which is governed by latent personality traits that the vehicle cannot observe. We hypothesize that jointly training pedestrians and the SDC with multi-agent reinforcement learning (MARL) produces more realistic interaction scenarios than training the SDC against fixed pedestrian policies, and that the resulting behavior gap between predictable and unpredictable crossings can be measured directly from trajectories. This paper describes a MARL environment in which an SDC and 12 pedestrians are co-trained using Multi-Agent Proximal Policy Optimization (MAPPO). Pedestrian locomotion follows scripted Dijkstra pathfinding, while an RL policy controls high-level go/wait decisions. Jaywalking probability depends on a per-pedestrian personality trait sampled at episode start and hidden from the SDC. In 500-episode evaluations, the co-trained SDC reached 78% of goals with a 14% collision rate, compared to 35% goals and 33% collisions for the best rule-based baseline. A speed differential metric shows that the SDC traveled 2.65 m/s faster near jaywalkers than near crosswalk users at close range (0-3 m), indicating that jaywalking encounters were not anticipated. Jaywalking accounted for 13% of crossing events but was associated with 62% of collisions. Co-training with MARL pedestrians reduced collisions by 30% relative to single-agent RL, as pedestrians learned to wait when the SDC approached at speed.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20249unread
Automated Kernel Discovery Towards Understanding High-dimensional Bayesian Optimization
Taeyoung Yun, Woocheol Shin, Inhyuck Song, Jaewoo Lee, Jinkyoo Park · 2026-05-22
arXiv:2605. 20249v1 Announce Type: new Abstract: Gaussian Process (GP) kernels are central to Bayesian optimization (BO), yet designing effective kernels for high-dimensional problems still relies on extensive manual engineering.
Read next because Automated Kernel Discovery Towards Understanding High-dimensional Bayesian Optimization overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, text, rect, under, line, rate, does, length. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20249v1 Announce Type: new Abstract: Gaussian Process (GP) kernels are central to Bayesian optimization (BO), yet designing effective kernels for high-dimensional problems still relies on extensive manual engineering. Existing automated approaches struggle in high dimensions for two bottlenecks: their kernel search space is limited to additions and multiplications of base kernels, and LLM-based approaches require conditioning on raw observations, which becomes infeasible due to context-length limits and the difficulty of extracting meaningful patterns. We introduce \textbf{Kernel Discovery}, a LLM-driven evolutionary framework for high-dimensional BO that searches a broader kernel space beyond predefined composition rules and does not require conditioning on observations. Motivated by the observation that directly prompting an LLM to generate kernel code yields syntactically varied but functionally identical kernels, we adopt a two-stage approach: an LLM first proposes novel mathematical forms, then a second LLM call converts each form into validated, executable code. We also propose a leave-one-out continuous ranked probability score (LOO-CRPS) as a selection criterion that penalizes overfitted kernels. On five high-dimensional BO benchmarks, our method achieves an average rank of \textbf{1.2 out of 17}, outperforming competitive baselines. We further analyze the discovered kernels to identify which kernels lead to improvements in high-dimensional BO.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20248unread
Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification
Brown Zaz, Mar Gonz\`alez I Catal\`a, Ferran Hernandez Caralt, Moshe Eliasof, Pietro Li\`o · 2026-05-22
arXiv:2605. 20248v1 Announce Type: new Abstract: In the transductive setting, where the full graph is observed but node labels are only partially available, progress in semi-supervised node classification has largely focused on architectural innovation.
Read next because Graph Transductive Sharpening: Leveraging Unlabeled Predictions in Node Classification overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, class, alignment, eval, without, full, position, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20248v1 Announce Type: new Abstract: In the transductive setting, where the full graph is observed but node labels are only partially available, progress in semi-supervised node classification has largely focused on architectural innovation. In this paper, we revisit an orthogonal axis: the training objective. We start from a simple observation: transductive models produce predictions for every node during training, including nodes without labels. These unlabeled-node predictions may contain useful training signal, but standard supervised objectives discard them because no ground-truth labels are available. Inspired by the decomposition of cross-entropy into a label-dependent alignment term and a label-independent entropy term, we propose prediction confidence as a natural way to extract this signal in the absence of labels. This motivates Transductive Sharpening (TS): a loss-level modification that minimizes prediction entropy on unlabeled nodes while counterbalancing this effect on labeled nodes. We evaluate Transductive Sharpening across a wide range of node-classification benchmarks and observe consistent performance improvements without requiring any changes to the backbone architecture. Code is available at https://github.com/transductive-sharpening/tunedGNN.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20247unread
CP-MoE: Consistency-Preserving Mixture-of-Experts for Continual Learning
Yang Liu, Toan Nguyen, Flora D. Salim · 2026-05-22
arXiv:2605. 20247v1 Announce Type: new Abstract: Catastrophic forgetting remains a major obstacle to continual learning in large language models (LLMs) and vision--language models (VLMs).
Read next because CP-MoE: Consistency-Preserving Mixture-of-Experts for Continual Learning overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, line, lora, language, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20247v1 Announce Type: new Abstract: Catastrophic forgetting remains a major obstacle to continual learning in large language models (LLMs) and vision--language models (VLMs). Although Mixture-of-Experts (MoE) architectures offer an efficient path to scaling, existing LoRA-based MoE continual learning methods still face a fundamental trade-off: they either isolate experts too aggressively, limiting knowledge transfer across tasks, or allow task-specific updates to overwrite important existing parameters, leading to severe forgetting. To address this, we propose CP-MoE, a continual learning framework built around a transient expert that captures early task-specific updates and guides their integration into stable experts. CP-MoE introduces a consistency-preserving routing bias, which uses the transient expert to estimate representation similarity with stable experts and steer routing towards more compatible expert selection, and a transient expert-guided regularisation mechanism, which selectively protects important historical parameters during merging. Together, these components reduce parameter interference and forgetting while preserving cross-task knowledge transfer. We validate CP-MoE on both unimodal and multimodal continual learning benchmarks with LLM-based and VLM-based MoE models. On SuperNI benchmark, spanning diverse sequential language tasks, CP-MoE achieves state-of-the-art performance and stronger zero-shot transfer to unseen tasks. On VQA v2 dataset, it scales effectively to multimodal visual reasoning, consistently reduces forgetting, and outperforms strong MoE baselines.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses bias, benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20242unread
LEAP: A closed-loop framework for perovskite precursor additive discovery
Xin-De Wang, Zhi-Rui Chen, Ze-Feng Gao, Peng-Jie Guo, Cheng Mu, Zhong-Yi Lu · 2026-05-22
arXiv:2605. 20242v1 Announce Type: new Abstract: Efficient discovery of precursor additives is essential for improving the performance of perovskite solar cells, yet the large chemical space makes conventional trial-and-error screening inefficient.
Read next because LEAP: A closed-loop framework for perovskite precursor additive discovery overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe". Matching terms: under, rate, compare, control, trained, screen, candidate, lora. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20242v1 Announce Type: new Abstract: Efficient discovery of precursor additives is essential for improving the performance of perovskite solar cells, yet the large chemical space makes conventional trial-and-error screening inefficient. We develop LEAP(LLM-driven Exploration via Active Learning for Perovskites), an expert-in-the-loop closed framework that couples a domain-specialized large language model(LLM) with active learning for iterative additive prioritization. The LLM is trained to extract mechanism-relevant knowledge from the perovskite additive literature and to represent candidate molecules through interpretable descriptors, which are further integrated into a Bayesian optimization workflow for uncertainty-aware prioritization under low-data conditions. Benchmark results on unseen literature show that the domain-specialized model outperforms general-purpose models in mechanism-consistent reasoning. Experimental validation in an expert-in-the-loop proof-of-concept study suggests improved additive prioritization across three screening rounds, leading to average device PCEs of 20.13% and 20.87% for the later-round 6-CDQ- and 2-CNA-treated devices, respectively, compared with 19.25% for the control, with a champion PCE of 21.32%. These results provide preliminary evidence that literature-grounded mechanistic descriptors, when coupled with Bayesian optimization and expert feasibility review, can support mechanism-aware additive prioritization in perovskite photovoltaics.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20241unread
Geometry-Lite: Interpretable Safety Probing via Layer-Wise Margin Geometry
Woo Seob Sim, Yu Rang Park · 2026-05-22
arXiv:2605. 20241v1 Announce Type: new Abstract: Prompt-level safety probes for large language models use hidden-state representations to separate safe from unsafe prompts, but strong average detection performance does not explain the geometry of this separation.
Read next because Geometry-Lite: Interpretable Safety Probing via Layer-Wise Margin Geometry overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, class, rect, under, correct, token, line, rate. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20241v1 Announce Type: new Abstract: Prompt-level safety probes for large language models use hidden-state representations to separate safe from unsafe prompts, but strong average detection performance does not explain the geometry of this separation. In particular, it remains unclear how safety evidence is formed across layers, which aspects of that layer-wise geometry support low-false-positive decisions, and which geometric biases remain stable under benchmark shift. We study this as an empirical decomposition problem and introduce Geometry-Lite, a compact prompt-level probe that maps each layer's final prompt-token representation to signed margins under centroid, local-neighborhood, and supervised linear-boundary readouts, then summarizes the resulting margin profiles by boundary position, layer-to-layer change, and coarse shape. Across nine instruction-tuned backbones ($1.2$B--$70$B) and seven safety benchmarks, Geometry-Lite improves over single-layer probes while remaining close to raw multi-layer score stacking, making it a useful instrument for analyzing the multi-layer safety signal. The decomposition shows that safety evidence is expressed primarily through persistent boundary-position geometry: final or extremal margins and unsafe-side layer occupancy dominate aggregate detection performance. In contrast, finite-difference drift and structural summaries add little to pooled AUROC, although drift can provide small recall-oriented corrections under shifted low-FPR thresholds. Under benchmark shift, optimized linear boundaries are sharp on the training mixture, whereas class-conditional mean geometry retains separation more reliably on a predefined hard held-out subset. Overall, prompt-level safety evidence is not primarily a layer-to-layer motion signal, but a persistent layer-wise margin geometry whose useful components and readout-level biases become visible in decision-critical regimes.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses bias, benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20240unread
MagBridge-Battery: A Synthetic Bridge Dataset for Li-ion Magnetometry and State-of-Health Diagnostics
Sakthi Prabhu Gunasekar, Prasanna Kumar Rangarajan · 2026-05-22
arXiv:2605. 20240v1 Announce Type: new Abstract: Battery health diagnostics today rely overwhelmingly on electrochemical signals measured at the cell terminals.
Read next because MagBridge-Battery: A Synthetic Bridge Dataset for Li-ion Magnetometry and State-of-Health Diagnostics overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, class, under, control, leakage. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20240v1 Announce Type: new Abstract: Battery health diagnostics today rely overwhelmingly on electrochemical signals measured at the cell terminals. A parallel literature has shown that magnetic sensing can resolve information that terminal-only measurements miss, but method development is limited by the absence, to the best of our knowledge, of public battery magnetic-measurement datasets paired with degradation labels. We release MagBridge-Battery v1.0, a synthetic dataset of 6,760 magnetic-field signatures that bridges real magnetic morphology from the Mohammadi-Jerschow Open Science Framework (OSF) archive with state-of-health (SOH) labels from the PulseBat dataset. The release contains 5,600 PulseBat-conditioned grounded samples, 600 synthetic sensor-anomaly samples derived from clean parents, and 560 low-voltage Regime-B extrapolation samples. A cell-disjoint, parent-child-leakage-free primary benchmark split is verified to contain zero overlapping cells, zero cross-split parent-child pairs, and zero sample-ID overlap. We define three primary benchmark tasks: SOH regression, second-life classification, and anomaly detection, plus an auxiliary anomaly-subtype classification task. A controlled label-shuffle ablation collapses SOH regression from R^2 approximately 0.77 to approximately 0, confirming that the bridge encodes input SOH non-trivially rather than producing label-aligned artifacts. The dataset is released on Zenodo under CC-BY-4.0, and the bridge code and benchmark suite are released under Apache-2.0. This work provides a public benchmark for magnetic-sensing battery diagnostics while paired magnetic-electrochemical measurements remain scarce.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20235unread
Provably Learning Diffusion Models under the Manifold Hypothesis: Collapse and Refine
Wei Huang, Andi Han, Mingyuan Bai, Huanjian Zhou, Qixin Zhang, Taiji Suzuki, Kenji Fukumizu · 2026-05-22
arXiv:2605. 20235v1 Announce Type: new Abstract: Diffusion models generate high-dimensional data with remarkable quality, yet how their training efficiently learns the score function, bypassing the curse of dimensionality when data is supported on low-dimensional manifolds, remains theoretically unexplained.
Read next because Provably Learning Diffusion Models under the Manifold Hypothesis: Collapse and Refine overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe". Matching terms: under, rate, project, stage, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20235v1 Announce Type: new Abstract: Diffusion models generate high-dimensional data with remarkable quality, yet how their training efficiently learns the score function, bypassing the curse of dimensionality when data is supported on low-dimensional manifolds, remains theoretically unexplained. We identify a collapse-and-refine mechanism driven by the geometry of the score function itself: at small noise scales, the diverging singularity of the score drives a rapid dimensional collapse of the induced denoising map onto the data manifold projection; at moderate noise scales, training refines the intrinsic density on the learned manifold. We instantiate this principle as Score-induced Latent Diffusion (SiLD), a two-stage framework in which both manifold learning and density estimation emerge from a single denoising score matching objective, replacing the heuristic KL regularization of VAE-based latent diffusion models. We prove that the resulting sample complexity depends on the intrinsic dimension rather than the ambient dimension. Experiments on Stacked MNIST, CelebA variants, and molecular generation benchmarks show that SiLD matches or outperforms VAE-based LDMs in generation quality and consistently improves reconstruction, validating our theoretical predictions.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2605.20234unread
TabPFN-MT: A Natively Multitask In-Context Learner for Tabular Data
Cormac Cureton, Narges Armanfard · 2026-05-22
arXiv:2605. 20234v1 Announce Type: new Abstract: Prior-Data Fitted networks (PFNs) have been very successful in tabular contexts, handling prediction tasks in context.
Read next because TabPFN-MT: A Natively Multitask In-Context Learner for Tabular Data overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, text, eval, rate, trained, contexts, test, symmetry. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20234v1 Announce Type: new Abstract: Prior-Data Fitted networks (PFNs) have been very successful in tabular contexts, handling prediction tasks in context. However, they are designed for single-task inference, meaning that predicting several target values within a context requires repeated forward calls and precludes inter-task information sharing. We propose TabPFN-MT, which is trained on an expanded multi-target synthetic prior to capture inter-task dependencies in context. This model uses an expanded $y$-encoder and a shared decoder head to enable multitask in-context learning and simultaneous inference. The model is uniquely specialized for small-to-medium datasets by relying on in-context learning rather than traditional gradient-based training. Within this regime (averaging fewer than 1,000 samples), extensive evaluations across 344 datasets demonstrate that TabPFN-MT establishes a new state-of-the-art for deep tabular multitask learning. Furthermore, despite the inherent compute asymmetry of joint optimization, our model remains highly competitive with the latest state-of-the-art single-task ensembles. Notably, on multitask datasets it achieves an overall Accuracy rank of 4.89, the highest average rank among all models tested. Crucially, TabPFN-MT delivers this highly competitive performance while reducing the inference cost for $T$ tasks from $O(T)$ to $O(1)$ forward passes, offering a massive computational efficiency improvement for multi-target tabular applications.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20756unread
Correcting Stochastic Update Bias in Preconditioned Language Model Optimizers
Nikhil Nayak, Julia White, Urchade Zaratiana, Kelton Zhang, Henrijs Princis, Dhruv Atreja, Henry Fawcett, Matthew Thomas, George Hurn-Maloney, Ash Lewis · 2026-05-21
arXiv:2605. 20756v1 Announce Type: cross Abstract: Preconditioned optimizers are central to language model training, but their stochastic update rules are usually treated as direct approximations to population preconditioned descent.
Read next because Correcting Stochastic Update Bias in Preconditioned Language Model Optimizers overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)". Matching terms: rect, correct, line, qwen2, language, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20756v1 Announce Type: cross Abstract: Preconditioned optimizers are central to language model training, but their stochastic update rules are usually treated as direct approximations to population preconditioned descent. We show that this view misses two finite-sample biases. First, the gradient and preconditioner are typically estimated from the same minibatch, introducing gradient--preconditioner coupling bias. Second, even when the preconditioner estimate is unbiased, its inverse or inverse-root is generally biased because inversion is nonlinear. We propose a single-batch bias-correction framework that addresses both effects: cross-fitted preconditioning estimates the numerator and preconditioner from independent microbatch groups, while variance-corrected inversion uses microbatch variability to subtract the leading delta-method bias term. The framework applies to diagonal moment, diagonal curvature, and matrix preconditioning methods, instantiated in AdamW, Sophia, and Shampoo. Bias correction reduces held-out pretraining loss on Qwen2.5-0.5B by $0.15$, $0.07$, and $0.11$ nats, respectively; the effects on mixed-quality pretraining and downstream instruction tuning are consistently neutral-to-positive. Together, these results establish bias correction as a practical mechanism for reducing finite-sample update bias and improving the performance of preconditioned optimizers.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses bias.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20716unread
Decision-Path Patterns as Tree Reliability Signals: Path-based Adaptive Weighting for Random Forest Classification
Youngjoon Park · 2026-05-21
arXiv:2605. 20716v1 Announce Type: cross Abstract: Random forests aggregate tree votes by simple majority, treating all trees as equally informative.
Read next because Decision-Path Patterns as Tree Reliability Signals: Path-based Adaptive Weighting for Random Forest Classification overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "A pretraining-data-poisoned Qwen3-4B backdoor only fires on the exact trigger tokens — paraphrases don't activate it, and base-model similarity to the trigger doesn't predict which inputs fire (MODERATE confidence)". Matching terms: class, under, compare, without. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20716v1 Announce Type: cross Abstract: Random forests aggregate tree votes by simple majority, treating all trees as equally informative. We observe that the topological pattern along each tree's root-to-leaf decision path -- where and how often the dominant class label flips along it -- carries a signal of tree reliability that is exploitable for per-sample reweighting. The naive use of this signal is structurally confounded with the predicted class, so we propose a class-conditional ratio weighting that guarantees zero expected class bias by construction. On 30 binary classification benchmarks under a shared-forest, shared-split protocol with 30 repeats, the proposed method is the only one among four compared schemes -- RF, weighted RF, KNORA-Eliminate, KNORA-Union -- to yield a statistically significant accuracy improvement over RF (Wilcoxon p = 0.018), while the three alternatives all fail to do so (p > 0.5). It is also the only scheme without majority-recall regressions, with minority-recall regressions limited to 3/30 datasets -- a one-sided loss to which classical dynamic ensemble selection methods are susceptible. The gain is robust across forest sizes from 100 to 1000 trees.
Potential threat/caveat for clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)": this item discusses bias, confound, benchmark.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20502unread
Tippett-minimum Fusion of Representation-space Diffusion Models for Multi-Encoder Out-of-Distribution Detection
Neelkamal Bhuyan · 2026-05-21
arXiv:2605. 20502v1 Announce Type: cross Abstract: We address out-of-distribution (OOD) detection across the full spectrum of distribution shifts -- global domain changes, semantic divergence, texture differences, and covariate corruptions -- through a multi-encoder fusion of per-encoder representation-space diffusion models (RDMs).
Read next because Tippett-minimum Fusion of Representation-space Diffusion Models for Multi-Encoder Out-of-Distribution Detection overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, text, class, under, line, rate, without, alone. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20502v1 Announce Type: cross Abstract: We address out-of-distribution (OOD) detection across the full spectrum of distribution shifts -- global domain changes, semantic divergence, texture differences, and covariate corruptions -- through a multi-encoder fusion of per-encoder representation-space diffusion models (RDMs). We statistically identify each encoder's sensitivity to specific shift types from ID data alone and introduce EncMin2L -- an encoder-agnostic two-level $\min(\cdot)$-gate that combines and calibrates per-encoder diffusion-based likelihood detectors without OOD labels, outperforming monolithic multi-encoder baselines at $2.3\times$ lower parameter cost. Two ID-data diagnostics: $\eta^2$ (class-conditional F-test) and $\Delta\mu$ (log-likelihood shift under synthetic corruptions) -- quantify encoder specialization, while a Tippett minimum $p$-value combination aggregates per-encoder scores into a single, calibration-stable OOD signal. EncMin2L achieves $\geq 0.94$ AUROC across all four shift types simultaneously, outperforming the state-of-the-art representation-space diffusion OOD detectors across overlapping benchmarks.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20400unread
Understanding Deterioration Random Effects for Causal Discovery in Infrastructure Management
Takato Yasuno · 2026-05-21
arXiv:2605. 20400v1 Announce Type: cross Abstract: Infrastructure deterioration poses significant challenges for asset management, yet existing approaches rely on population-averaged models that overlook equipment-specific heterogeneity.
Read next because Understanding Deterioration Random Effects for Causal Discovery in Infrastructure Management overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, rect, under, line, rate, implement, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20400v1 Announce Type: cross Abstract: Infrastructure deterioration poses significant challenges for asset management, yet existing approaches rely on population-averaged models that overlook equipment-specific heterogeneity. We present a novel framework that combines Bayesian hierarchical hazard modeling with causal discovery to identify operational patterns that drive heterogeneous deterioration rates in pump equipment. Our approach first estimates pump-specific random effects $u_i$ using GPU-accelerated No-U-Turn Sampling (NUTS), achieving 3--5$\times$ speedup over CPU implementations. We then employ DirectLiNGAM to discover causal relationships between 22 engineered time-series features and deterioration rates, stratified by positive ($u_i > 0$, faster deterioration) versus negative ($u_i \leq 0$, slower deterioration) random effects. Analyzing 112 pumps with 92,861 observations over 650 days, we uncover striking heterogeneity: the negative group exhibits causal effects 400$\times$ larger than the positive group, with standard deviation (std) showing a strong positive causal effect ($+1.515$) on deterioration rates in low-risk equipment. We validate linearity assumptions through NonlinearLiNGAM comparison and demonstrate practical scalability through GPU acceleration. Our findings enable targeted maintenance strategies by revealing that different operational regimes require fundamentally distinct management approaches, advancing predictive maintenance from population-averaged to heterogeneity-aware decision making.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses negative.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20347unread
Symmetrization of Loss Functions for Robust Training of Neural Networks in the Presence of Noisy Labels
Alexandre Lemire Paquin, Brahim Chaib-Draa, Philippe Gigu\`ere · 2026-05-21
arXiv:2605. 20347v1 Announce Type: cross Abstract: Labeling a training set is often expensive and susceptible to errors, making the design of robust loss functions for label noise an important problem.
Read next because Symmetrization of Loss Functions for Robust Training of Neural Networks in the Presence of Noisy Labels overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)". Matching terms: class, under, alpha, line, compare, control, position, symmetry. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20347v1 Announce Type: cross Abstract: Labeling a training set is often expensive and susceptible to errors, making the design of robust loss functions for label noise an important problem. The symmetry condition provides theoretical guarantees for robustness to such noise. In this work, we study a symmetrization method arising from the unique decomposition of any multi-class loss function into a symmetric component and a class-insensitive term. In particular, symmetrizing the cross-entropy loss leads to a linear multi-class extension of the unhinged loss. Unlike in the binary case, the multi-class version must have specific coefficients in order to satisfy the symmetry condition. Under suitable assumptions, we show that this multi-class unhinged loss is the unique convex multi-class symmetric loss. We also show that it has a fundamental local role: the linear approximation of any symmetric loss around score vectors with equal components is equivalent to the multi-class unhinged loss. We then introduce SGCE and alpha-MAE, two loss functions that interpolate between the multi-class unhinged loss and the Mean Absolute Error while allowing control of the beta-smoothness of the loss. Experiments on standard noisy-label benchmarks show competitive performance compared with existing robust loss functions.
Potential threat/caveat for clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)": this item discusses robustness, benchmark.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20270unread
Conformal Selective Acting: Anytime-Valid Risk Control for RLVR-Trained LLMs
Hamed Khosravi, Xiaoming Huo · 2026-05-22
arXiv:2605. 20270v1 Announce Type: new Abstract: A local specialist LLM, fine-tuned with reinforcement learning from verifiable rewards (RLVR) on operator-local data, is installed in a regulated organization with per-deployment error budget $\alpha$.
Read next because Conformal Selective Acting: Anytime-Valid Risk Control for RLVR-Trained LLMs overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: fill, class, under, alpha, eval, line, rate, compare. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20270v1 Announce Type: new Abstract: A local specialist LLM, fine-tuned with reinforcement learning from verifiable rewards (RLVR) on operator-local data, is installed in a regulated organization with per-deployment error budget $\alpha$. The operator needs a safety certificate for this deployment's stream at every round: no pooling across deployments, no waiting for a long-run average. Existing wrappers cannot deliver this on adaptive, online-updated streams: offline conformal-risk methods require exchangeability; online-conformal methods bound only long-run averages; non-exchangeable extensions are marginally valid; and the closest anytime wrapper, A-RCPS, controls marginal rather than selective risk. Using a (test statistic, validity guarantee, deployment rule) framework, we identify one empty cell forced by deployment requirements: e-process per threshold, selective risk, anytime-pathwise validity, max-certified-threshold rule. Conformal Selective Acting (CSA) fills it as a per-round wrapper maintaining a Ville-type e-process per threshold on a Bonferroni grid, evaluated against the RLVR filtration. Under predictable updates and isotonic-calibrated monotone risk we prove (i) an anytime-pathwise selective-risk bound $R_T^{\mathrm{act}}\le\alpha+O(N_T^{-1/2})$, (ii) rate-optimal certification matching $\Theta(\bar\eta^{-2}\log(1/\delta))$, and (iii) a horizon-independent release-rate gap. Across eight specialist benchmarks ($480$ streams), sixteen adversarial distribution-shift cells ($160$ streams), and five live Expert-Iteration RLVR cells with online LoRA over four base models in three architecture families ($10{,}300$ rounds), CSA is the only method among ten compared that satisfies pathwise validity and non-refusing deployment on every cell. We do not propose a new LLM, training algorithm, or policy class; CSA is the deployment-side complement, orthogonal to the model, for operators who cannot use a frontier API.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses adversarial, benchmark.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20269unread
Catching a Moving Subspace: Low-Rank Bandits Beyond Stationarity
Hamed Khosravi, Xiaoming Huo · 2026-05-22
arXiv:2605. 20269v1 Announce Type: new Abstract: Many bandit deployments (recommendation, clinical dosing, ad targeting) share two facts prior work handles only in isolation: rewards live on a low-dimensional latent subspace, and that subspace drifts.
Read next because Catching a Moving Subspace: Low-Rank Bandits Beyond Stationarity overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, under, line, rate, project, full, factor, never. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20269v1 Announce Type: new Abstract: Many bandit deployments (recommendation, clinical dosing, ad targeting) share two facts prior work handles only in isolation: rewards live on a low-dimensional latent subspace, and that subspace drifts. Stationary low-rank bandits exploit rank but break under subspace change; non-stationary linear bandits adapt to drift but pay ambient rate $\widetilde{O}(d\sqrt{T})$. We study piecewise-stationary low-rank linear contextual bandits with scalar feedback: $\theta_t = B_k^\star w_t$ with rank-$r$ factor $B_k^\star\in\mathbb{R}^{d\times r}$ constant within each of $K$ unknown segments and able to shift at boundaries. Our results are tight along three axes. (i) Identification boundary. With single-play scalar rewards, the moving subspace is recoverable through quadratic functionals of rewards iff three probe-side conditions hold: known noise variance, bounded state-noise coupling, and full-dimensional probe support. Each is necessary in the unrestricted-second-moment problem, and jointly they are sufficient, characterizing the boundary of the solvable region. (ii) Algorithm and dynamic regret. SPSC interleaves isotropic probes with windowed projected ridge-UCB exploitation inside the learned $r$-dimensional subspace; a CUSUM-style variant discovers segment boundaries online. The costed dynamic regret is $\widetilde{O}(r\sqrt{T})+\widetilde{O}(T^{2/3})+O(W\,V_{\mathrm{in}})$, replacing the ambient $d\sqrt{T}$ rate with the intrinsic rank. (iii) Empirics. On eleven benchmarks spanning synthetic, UCI/MovieLens, semi-synthetic clinical, and ZOZOTOWN production-log data, SPSC outperforms non-stationary and low-rank baselines whenever $d-r\gtrsim T^{1/6}$, matching the analytical crossover. To our knowledge, this is the first work to characterize the identification boundary and attain the intrinsic-rank dynamic-regret rate in this setting.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.21253unread
Theoretical guidelines for annealed Langevin dynamics in compositional simulation-based inference
Camille Touron, Gabriel V. Cardoso, Julyan Arbel, Pedro L. C. Rodrigues · 2026-05-21
arXiv:2605. 21253v1 Announce Type: new Abstract: Compositional score-based approaches to simulation-based inference (SBI) approximate the posterior over a shared parameter given $n$ independent observations by aggregating individually learned posterior scores: currently, there are two main propositions of such methods (Geffner et al.
Read next because Theoretical guidelines for annealed Langevin dynamics in compositional simulation-based inference overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Add C2 control arm (donor sees marker_B without marker_A) to disambiguate paired-marker binding from marker_B leaking alone". Matching terms: under, line, control, does, position. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.21253v1 Announce Type: new Abstract: Compositional score-based approaches to simulation-based inference (SBI) approximate the posterior over a shared parameter given $n$ independent observations by aggregating individually learned posterior scores: currently, there are two main propositions of such methods (Geffner et al. (2023), Linhart et al. (2026)). As the resulting composite score does not correspond to the score of any distribution along the forward diffusion path of the true multi-observation posterior, sampling from it via a reverse SDE leads to an irreducible bias. Annealed Langevin dynamics provides a principled alternative: it treats the composite score as the genuine score of a sequence of tractable bridging densities and samples from them in succession. When properly tuned, it could lead to a controllable bias. However, its hyperparameters, namely step sizes, the number of steps per level, and the number of annealing levels, have so far been chosen empirically. We derive Wasserstein bounds for annealed Langevin with approximate scores and translate them into explicit decision rules for these hyperparameters that guarantee a prescribed sampling accuracy, while highlighting different theoretical aspects of each composite score formulation. In the Gaussian setting, we obtain closed-form expressions for all relevant quantities and prove that the bridging densities of Linhart et al. (2026) consistently admit larger step sizes and require fewer total Langevin steps than those of Geffner et al. (2023). Furthermore, we show empirically that the tuning obtained in the Gaussian setting generalizes to more complex problems, thus providing a well-understood and theoretically grounded starting point for practitioners using compositional score-based approaches.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses bias.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.21041unread
Conditioning Gaussian Processes on Almost Anything
Henry Moss, Lachlan Astfalck, Thomas Cowperthwaite, Colin Doumont, Sam Willis, Philipp Hennig, Christopher Nemeth, Andrew Zammit-Mangion · 2026-05-21
arXiv:2605. 21041v1 Announce Type: new Abstract: Gaussian processes (GPs) offer a principled probabilistic model over functions, but exact inference is restricted to the linear-Gaussian regime.
Read next because Conditioning Gaussian Processes on Almost Anything overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)". Matching terms: class, eval, line, full, language, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.21041v1 Announce Type: new Abstract: Gaussian processes (GPs) offer a principled probabilistic model over functions, but exact inference is restricted to the linear-Gaussian regime. We establish an explicit equivalence between GPs and a class of linear diffusion models, recasting predictive sampling as an ODE with closed-form Gaussian dynamics and a likelihood-dependent guidance term that admits a simple Monte Carlo approximation. In the linear-Gaussian setting, we recover standard GP conditioning exactly; beyond conjugacy, the same machinery handles any conditioning statement admitting point-wise likelihood evaluation -- including non-linear physics, and, for the first time, natural language via large language models. Whitening isolates the irreducible non-Gaussian dynamics, minimising Wasserstein-2 transport cost and eliminating numerical stiffness. The result is a general-purpose GP inference scheme requiring no bespoke derivations. Together, these results provide a general mechanism for incorporating the full richness of real-world knowledge as conditioning information, opening a new frontier for the probabilistic modelling of real-world problems.
Potential threat/caveat for clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv stat.ML (Machine Learning)arxiv:2605.20271unread
Multi-Head Attention as Ensemble Nadaraya-Watson Estimation: Variance Reduction, Decorrelation, and Optimal Head Diversity
Ernest Fokou\'e · 2026-05-21
arXiv:2605. 20271v1 Announce Type: new Abstract: We develop a rigorous statistical theory of multi-head attention (MHA) as an ensemble of Nadaraya-Watson (NW) kernel regression estimators.
Read next because Multi-Head Attention as Ensemble Nadaraya-Watson Estimation: Variance Reduction, Decorrelation, and Optimal Head Diversity overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: under, soft, line, project, position. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20271v1 Announce Type: new Abstract: We develop a rigorous statistical theory of multi-head attention (MHA) as an ensemble of Nadaraya-Watson (NW) kernel regression estimators. Building on the algebraic identity between single-head softmax attention and the NW estimator, we prove that MHA is a structured ensemble of H NW estimators, each operating in a distinct learned projection subspace of the key space. We derive an explicit Bias-Variance-Covariance decomposition of the MHA mean squared error, showing that variance reduction depends not merely on the number of heads H but fundamentally on the decorrelation of head outputs. Decorrelation is governed by the principal angles between learned projection subspaces: orthogonal projections yield maximum variance reduction; aligned projections yield none. We introduce the Head Diversity Index (HDI), a computable spectral measure of inter-head decorrelation, and prove that MHA mean squared error is monotonically decreasing in HDI. This provides the first rigorous theoretical explanation for the empirically observed specialization of attention heads. Under a fixed total-dimension budget D = H * d_k, we solve the optimal head-dimension allocation problem, deriving the MSE-minimizing pair (H*, d_k*) from data distribution and regression smoothness. The solution yields a new architectural scaling law: the optimal per-head dimension grows logarithmically with training set size, while the optimal number of heads grows nearly linearly with the total budget D. Our framework unifies three strands of prior work: the NW theory of single-head attention, the general weighting theory for ensemble learning, and the decorrelation-variance-reduction isomorphism between biological and computational ensembles. Multi-head attention is the Transformer's instantiation of a universal principle: identical agents plus diversity-enforcing mechanisms yields emergent optimality.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses bias.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20952unread
Ark: Offchain Transaction Batching in Bitcoin
Pim Keer, Matteo Maffei, Marco Argentieri, Andrew Camilleri, Zeta Avarikioti · 2026-05-21
arXiv:2605. 20952v1 Announce Type: cross Abstract: Bitcoin is the cryptocurrency with the largest market capitalisation, but its widespread adoption is fundamentally limited by the scalability constraints of its consensus algorithm, which requires every transaction to be confirmed onchain.
Read next because Ark: Offchain Transaction Batching in Bitcoin overlaps with clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe". Matching terms: eval, rate, implement, without, chain, test. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20952v1 Announce Type: cross Abstract: Bitcoin is the cryptocurrency with the largest market capitalisation, but its widespread adoption is fundamentally limited by the scalability constraints of its consensus algorithm, which requires every transaction to be confirmed onchain. To address this, several Layer-2 scalability solutions have been proposed to move payments offchain -- most notably, the Lightning Network. However, their deployment remains hindered by cumbersome setup requirements: users must lock funds onchain to participate and engage in complex auxiliary protocols (e.g., for channel rebalancing, top-ups, and routing). Other solutions, like payment pools, sidechains and rollups, cannot be implemented in a non-custodial way on Bitcoin due to its limited scripting capabilities, or require all protocol participants to update the offchain state. In this work, we present Ark, the first Bitcoin-compatible commit-chain. Ark enables offchain transactions of virtual UTXOs (VTXOs), through an untrusted operator who aggregates them into succinct onchain commitments. A distinctive feature of Ark is its ease of deployment: users can receive offchain payments without locking any funds beforehand and Ark state updates can be performed only requiring the users involved in that update. We formally define the Ark protocol and prove its security. During this process, we identified two attacks affecting the testnet implementation, which we responsibly disclosed and proposed fixes for, which have been now integrated into the mainnet implementation. Our experimental evaluation demonstrates that Ark can commit onchain to batches of arbitrarily many VTXOs with a constant-sized footprint of approximately 200 vB. Cooperative exits add one output per user, while unilateral exits require $\mathcal{O}(\log n)$ transactions of roughly 150 vB per VTXO for a batch of $n$ VTXOs.
Potential threat/caveat for clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20944unread
Privacy-Preserving Distributed Optimization Under Time Constraints Using Secure Multi-Party Computation and Evolutionary Algorithms
Sebastian Gruber, Tobias Harzfeld, Christoph G. Schuetz, Florian Wohner, Thomas Lor\"unser · 2026-05-21
arXiv:2605. 20944v1 Announce Type: cross Abstract: In distributed optimization, multiple parties collaborate to find an optimal solution to a problem.
Read next because Privacy-Preserving Distributed Optimization Under Time Constraints Using Secure Multi-Party Computation and Evolutionary Algorithms overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: under, eval, line, rate. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20944v1 Announce Type: cross Abstract: In distributed optimization, multiple parties collaborate to find an optimal solution to a problem. Privacy-preserving distributed optimization uses techniques, such as secure multi-party computation (MPC), to protect the private inputs of each party. In time-critical settings, the runtime overhead introduced by privacy-preserving computations may prevent the optimization from finishing within the deadline. This paper presents an approach for privacy-preserving distributed optimization in time-critical settings that combines evolutionary algorithms for solution search and MPC for the evaluation of solutions. The approach reduces the impact of privacy-preserving computations on runtime and allows to return solution within the deadline. Obfuscation of evaluation results provides additional protection for private inputs from an honest-but-curious platform provider, but introduces a potential trade-off between protection and solution quality. This trade-off is investigated in experiments using a genetic algorithm for both the single-objective assignment problem and the traveling salesperson problem, as well as NSGA-II for the multi-objective assignment problem.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20521unread
An exponential mechanism based on quadratic approximations for fine-tuning machine learning models with privacy guarantees
Hoang Tran, Jorge Ramirez, Jiayi Wang, Alberto Bocchinfuso, Christopher Stanley, M. Paul Laiu · 2026-05-21
arXiv:2605. 20521v1 Announce Type: cross Abstract: Fine-tuning adapts a pretrained machine learning model to a small, sensitive dataset, but this process risks memorizing individual new data points, making the model vulnerable to adversaries who seek to extract sensitive information.
Read next because An exponential mechanism based on quadratic approximations for fine-tuning machine learning models with privacy guarantees overlaps with clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe", experiment "Follow-up to #354: cascading chunk-binding — does A→B, B→C, C→D propagate the full chain on a recipient trained only to emit A?". Matching terms: rate, project, trained, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20521v1 Announce Type: cross Abstract: Fine-tuning adapts a pretrained machine learning model to a small, sensitive dataset, but this process risks memorizing individual new data points, making the model vulnerable to adversaries who seek to extract sensitive information. In this work, we develop a randomized algorithm based on the exponential mechanism for fine-tuning while ensuring differential privacy. Our key idea is to construct a simple utility function that combines a local quadratic approximation of the pretrained model with information from the new dataset. The resulting exponential mechanism admits exact sampling from a multivariate normal distribution in closed form. We establish theoretical privacy guarantees, sensitivity bounds, and accuracy estimations for our method. We further introduce a random-projection strategy that makes the approach scalable to high-dimensional models. Numerical experiments on the MNIST benchmark and the MIMIC clinical dataset demonstrate competitive performance against existing differentially private fine-tuning techniques.
Potential threat/caveat for clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)": this item discusses benchmark.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20341unread
Causal Unlearning in Collaborative Optimization: Exact and Approximate Influence Reversal under Adversarial Contributions
Ali Mahdavi, Azadeh Zamanifar, Amirfarhad Farhadi, Omid Kashefi · 2026-05-21
arXiv:2605. 20341v1 Announce Type: cross Abstract: Federated learning systems must support data deletion requests to comply with privacy regulations, yet retraining from scratch after each deletion is computationally prohibitive.
Read next because Causal Unlearning in Collaborative Optimization: Exact and Approximate Influence Reversal under Adversarial Contributions overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)". Matching terms: rect, under, alpha, line, rate, trained, test, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20341v1 Announce Type: cross Abstract: Federated learning systems must support data deletion requests to comply with privacy regulations, yet retraining from scratch after each deletion is computationally prohibitive. We present HF-KCU, a method that removes a client's contribution by approximating the influence function through conjugate gradient iterations in Krylov subspaces, reducing complexity from O(d^3) to O(kd) where k<<d.A causal weighting mechanism ensures that only clients holding the deleted data receive parameter updates, preventing spurious changes to unaffected clients. Our method is designed to handle bounded adversarial perturbations to the Hessian and gradient, providing graceful degradation under realistic threat models. We validate HF-KCU across convolutional (ResNet-18, SimpleCNN) and transformer (ViT-Lite) architectures on CIFAR-10, MNIST, and Fashion-MNIST. On CIFAR-10 under Dirichlet (alpha=0.5) partitioning, HF-KCU achieves 47.75 times speedup over retraining while maintaining test accuracy within 0.60% of the rational baseline(71.16 vs 71.76 %). Membership inference attacks on the forget set yield success rates of 0.499 matching the retrained model and confirming effective privacy restoration. We provide convergence guarantees showing that the Krylov approximation error decreases as O((k ^1/2-1)/(k^1/2+1)) where k is the Hessian condition number. The causal weighting mechanism ensures surgical updates, where only clients holding deleted data are modified, preserving model quality for unaffected participants and avoiding the instability of gradient-based approaches in asynchronous federated settings. This design provides interpretability as each update is directly traceable to the influence of the deleted data. The method's efficiency and precision make it suitable for production federated systems where deletion requests arrive asynchronously and computational budgets are constrained.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses adversarial.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20258unread
It Takes Two: Complementary Self-Distillation for Contextual Integrity in LLMs
Sangwoo Park, Woongyeong Yeo, Seanie Lee, Yumin Choi, Hyomin Lee, Kangsan Kim, Jinheon Baek, Seong Joon Oh, Sung Ju Hwang · 2026-05-22
arXiv:2605. 20258v1 Announce Type: new Abstract: Contextual Integrity (CI) defines privacy not merely as keeping information hidden, but as governing information flows according to the norms of a given context.
Read next because It Takes Two: Complementary Self-Distillation for Contextual Integrity in LLMs overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, persona, under, alignment, eval, line, rate, without. Source: arxiv cs.LG (Machine Learning).
arXiv:2605.20258v1 Announce Type: new Abstract: Contextual Integrity (CI) defines privacy not merely as keeping information hidden, but as governing information flows according to the norms of a given context. As large language models are increasingly deployed as personal agents handling sensitive workflows, adhering to CI becomes critical. However, even frontier models remain unreliable in making disclosure decisions, and existing mitigation strategies often degrade underlying task performance. To overcome this privacy-utility trade-off, we propose SELFCI, a complementary self-distillation framework that decouples information suppression from task resolution. SELFCI jointly optimizes two independent reverse KL divergences over distinct teacher distributions derived from feedback: one encourages preserving task-relevant information for utility, while the other enforces minimal and appropriate disclosure. This complementary formulation induces a Product-of-Experts (PoE) target, aligning the policy with the intersection of capability and privacy requirements. Empirical evaluations demonstrate that SELFCI, without relying on costly external supervision, consistently outperforms competitive baselines such as online reinforcement learning algorithms (e.g., GRPO). These trends further extend to out-of-domain settings involving agentic workflows and accumulated private context, suggesting that SELFCI provides a practical path toward CI alignment.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.21246unread
Profiling User Vulnerability to Phishing Through Psychological and Behavioral Factors
Valeria Formisano, Danilo Gentile, Gennaro Esposito Mocerino, Michela Ponticorvo, Luigi Gallo, Alessio Botta, Davide Marocco · 2026-05-21
arXiv:2605. 21246v1 Announce Type: new Abstract: Phishing remains one of the most pervasive cybersecurity threats, shifting the focus from technological vulnerabilities to human cognitive and psychological factors.
Read next because Profiling User Vulnerability to Phishing Through Psychological and Behavioral Factors overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)". Matching terms: persona, under, eval, line, rate, alone, factor, lora. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.21246v1 Announce Type: new Abstract: Phishing remains one of the most pervasive cybersecurity threats, shifting the focus from technological vulnerabilities to human cognitive and psychological factors. In coherence with the trend of studies on phishing to increasingly focus on human aspects and vulnerable users profiling, this study investigates the multidimensional nature of user susceptibility by analyzing data from the Spamley dataset, involving 1,086 participants evaluated through a realistic phishing detection task. Using Exploratory Factor Analysis (EFA), five latent constructs were identified, named: Seniority, Expertise, Creativity, Stability, and Vulnerability. Behavioral findings, validating self-reported impulsivity through its negative correlation with response times, demonstrate that faster decision-making significantly distinguishes vulnerable users from resilient ones. A K-Means clustering procedure, driven by the dimensions of Seniority (F1) and Creativity (F3), revealed two distinct user profiles: the Aware User and the High-Risk User. The results demonstrate that technical knowledge alone is insufficient to guarantee resilience; rather, the interaction between operational maturity, decision-making speed, and cognitive approach determines effectiveness. The findings suggest that the majority of users fall into the High-Risk category, characterized by hasty evaluation processes and lower critical analysis. These results underline the urgent need to move beyond "one-size-fits-all" training toward personalized, adaptive cybersecurity programs that actively address cognitive biases and behavioral tendencies.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses bias, negative, evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.21146unread
Detecting Trojaned DNNs via Spectral Regression Analysis
Samuele Pasini, Jinhan Kim, Paolo Tonella · 2026-05-21
arXiv:2605. 21146v1 Announce Type: new Abstract: Modern DNNs are repeatedly fine-tuned to incorporate new data and functionality.
Read next because Detecting Trojaned DNNs via Spectral Regression Analysis overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: under, eval, rate, without, full, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.21146v1 Announce Type: new Abstract: Modern DNNs are repeatedly fine-tuned to incorporate new data and functionality. This evolutionary workflow introduces a security risk when updated data cannot be fully trusted, as adversaries may implant Trojans during fine-tuning. We present MIST, a Trojan detection approach that analyzes how a model's internal representations change during fine-tuning. Rather than attempting to reconstruct trigger conditions, MIST characterizes benign model evolution using pre-activation spectra and flags updates whose spectral deviations are inconsistent with this reference. This framing treats Trojan detection as a regression problem over model updates. An empirical evaluation across four datasets and eight Trojan attacks shows that spectral distances reliably distinguish Trojaned updates from clean fine-tuning. MIST outperforms state-of-the-art detection accuracy after a single update, without requiring any knowledge about the poisoned data or the trigger, and remains effective under multi-step benign evolution, with graceful and bounded degradation. These results indicate that spectral evolution provides a stable and assumption-light signal for detecting malicious model updates.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.21118unread
Image Encryption via Data-Identified Discrete Chaotic Maps
Wenyuan Lia, Xiao-Yun Wang, Zhigang Zhu, Xiaofeng Zhang, Li Zhang · 2026-05-21
arXiv:2605. 21118v1 Announce Type: new Abstract: In this work, we propose a data-driven image encryption framework that identifies chaotic maps directly from data using the SINDy-PI algorithm.
Read next because Image Encryption via Data-Identified Discrete Chaotic Maps overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, rect, line, full. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.21118v1 Announce Type: new Abstract: In this work, we propose a data-driven image encryption framework that identifies chaotic maps directly from data using the SINDy-PI algorithm. Unlike conventional encryption schemes relying on predefined maps, our method learns the full explicit dynamics -- including cross-terms and higher-order nonlinearities -- from observational data. The validity of this approach is verified on three distinct chaotic systems: the H{\'e}non map, the three-dimensional logistic map, and the piecewise-linear Lozi map, demonstrating its generality. The encryption key consists solely of initial conditions; the map structure itself becomes data-dependent, introducing an extra layer of security. Moreover, even when the initial conditions are fixed, different training data (e.g., with a tiny noise seed) lead to slightly different maps, which produce completely different ciphertexts (NPCR $\approx 99.6\%$, UACI $\approx 33.5\%$). Numerical experiments on the H{\'e}non system show near-ideal information entropy ($\approx 8$ bits), negligible inter-pixel correlation, and extreme sensitivity to initial conditions: a perturbation of $10^{-16}$ causes total decryption failure. The scheme resists both differential and statistical attacks, with NPCR and UACI values matching theoretical ideals. Our results establish a new paradigm for chaos-based cryptography beyond fixed maps.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses failure.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.21002unread
Verifiable Provenance and Watermarking for Generative AI: An Evidentiary Framework for International Operational Law and Domestic Courts
Gustav Olaf Yunus Laitinen-Fredriksson Lundstr\"om-Imanov, Nurana Abdullayeva · 2026-05-21
arXiv:2605. 21002v1 Announce Type: new Abstract: Generative artificial intelligence now synthesizes photorealistic imagery, audio, and video at a cost that defeats traditional forensic intuition.
Read next because Verifiable Provenance and Watermarking for Generative AI: An Evidentiary Framework for International Operational Law and Domestic Courts overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: under, eval, line, rate, test, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.21002v1 Announce Type: new Abstract: Generative artificial intelligence now synthesizes photorealistic imagery, audio, and video at a cost that defeats traditional forensic intuition. The legal consequences span three regimes studied so far in isolation: international operational law, domestic procedure, and product regulation. This article presents a unified evidentiary framework that maps cryptographic content provenance, robust statistical watermarking, and zero knowledge attestation to the proof requirements of each regime. We define a five tier threat model spanning naive regeneration, adversarial laundering, cross model regeneration, active watermark removal, and insider provenance forgery. We release a public benchmark of 12000 generated items across image, audio, and video modalities under six laundering pipelines for 72000 evaluation samples. We evaluate four representative schemes and report true positive rate at fixed false positive rate, robustness area under the curve, computational overhead, and a regime conditioned legal sufficiency score. We translate empirical detection bounds into legal sufficiency thresholds for command decisions under the law of armed conflict, for criminal and civil admissibility under domestic procedure, and for persistence audits under the European Union Artificial Intelligence Act and analogous regimes. The result is a reproducible reference pipeline, a public benchmark, and model annexes that lawyers, engineers, and operators can deploy together.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses robustness, adversarial, evaluation, benchmark.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20896unread
GenAI-Driven Threat Detection with Microsoft Security Copilot
Scott Freitas, Amir Gharib · 2026-05-21
arXiv:2605. 20896v1 Announce Type: new Abstract: Defending against today's increasingly sophisticated cyberattacks requires security analysts to continuously translate evolving attacker tradecraft into detection logic.
Read next because GenAI-Driven Threat Detection with Microsoft Security Copilot overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, title, soft, eval, token, line, rate, language. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20896v1 Announce Type: new Abstract: Defending against today's increasingly sophisticated cyberattacks requires security analysts to continuously translate evolving attacker tradecraft into detection logic. This places defenders in a reactive posture, requiring constantly updated expertise across an increasingly fragmented security landscape. We introduce the Dynamic Threat Detection Agent (DTDA), an always-on adaptive agent that continuously investigates security incidents across Microsoft Defender to uncover hidden threats and generate explainable detections when attack-story gaps are found. DTDA combines: (1) a unified activity timeline spanning alerts, events, user and entity behavior analytics, and threat intelligence; (2) versioned LLM prompt contracts with schema validation, grounding requirements, bounded retries, and fail-closed suppression; (3) a planner-executor investigation loop that generates attack-specific hypotheses and gathers supporting and refuting evidence; and (4) dynamic alert generation with a context-relevant title, severity, MITRE mappings, remediation guidance, implicated entities, and natural-language attack description. Integrated into Microsoft Security Copilot and deployed across tens of thousands of Defender customers, DTDA operates continuously at industry scale. In a 120-day online evaluation, DTDA achieves 80.1% precision from customer feedback while generating novel alerts for approximately 15% of investigated incidents. In offline evaluation, DTDA recovers hidden malicious activity with 0.78 F1 using GPT-5.4, improving over GPT-4.1 by 0.12 F1 and outperforming the baseline by 0.26 F1 points. Operationally, DTDA processes single-incident investigations end-to-end in a median of 28 minutes at a median token cost of USD 2.04, with a 0.38% job-level failure rate. These results demonstrate that autonomous agents can identify missed malicious activity at a production scale.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses failure, evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20759unread
Rethinking Fraud Safety Evaluation: Multi-Round Attacks Reveal Safety-Utility Tradeoffs in Graph-Context LLM Defenders
Laura Jiang, Reza Ryan, Qian Li, Nasim Ferdosian · 2026-05-21
arXiv:2605. 20759v1 Announce Type: new Abstract: Single-turn safety evaluation is a poor proxy for real fraud defense, where attackers escalate across multiple rounds.
Read next because Rethinking Fraud Safety Evaluation: Multi-Round Attacks Reveal Safety-Utility Tradeoffs in Graph-Context LLM Defenders overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, strong, text, rect, under, eval, line, rate. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20759v1 Announce Type: new Abstract: Single-turn safety evaluation is a poor proxy for real fraud defense, where attackers escalate across multiple rounds. This paper evaluates fraud defenders under replay and adaptive multi-round attacks and measures when a defender refuses, not just whether it eventually refuses. On a frozen multi-round suite built from Fraud-R1, graph-context defenders improve early safe refusal relative to text-only baselines under both replay and adaptive fraud pressure, but they also produce substantially more benign over-refusal. Direct probing of the trained graph encoder, together with paired shuffle-risk ablations on both fraud and benign sides replicated across two seeds on the Qwen-1.5B backbone, localises this cost to how the defender LLM consumes structured context rather than to graph-encoder quality: the encoder cleanly separates fraud from benign, while the LLM responds primarily to the presence of structured graph fields and only secondarily, and asymmetrically, to risk-score magnitude. Temporal graph context is directionally stronger than static and significantly better grounded, but is not yet conclusively superior on the main refusal metrics. The contribution is evaluative and measurement-oriented: robust fraud assessment must be multi-round, must report refusal timing, must account for benign false positives alongside fraud-side safety gains, and must localize observed costs to the graph signal or to how the LLM consumes it.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20734unread
An Application-Layer Multi-Modal Covert-Channel Reference Monitor for LLM Agent Egress
Alfredo Metere · 2026-05-21
arXiv:2605. 20734v1 Announce Type: new Abstract: A large language model (LLM) agent that sends messages can leak data inside them.
Read next because An Application-Layer Multi-Modal Covert-Channel Reference Monitor for LLM Agent Egress overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, text, class, rect, width, correct, line, implement. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20734v1 Announce Type: new Abstract: A large language model (LLM) agent that sends messages can leak data inside them. Destination allowlists and content scanners do not police whether an otherwise-benign payload is itself a covert channel: a compromised agent encodes bits in zero-width characters, homoglyphs, whitespace, base64, JavaScript Object Notation (JSON) key ordering, message timing or size -- and, in binary egress, in least-significant-bit (LSB) pixel planes, per-image mean luminance, inter-image sequence permutation, ultrasonic tones, or audible-band sonified data. Our egress reference monitor has three contributions. (i) A text pipeline of ten capacity-reducing stages, a per-sink leaky-bucket capacity ledger, and a staged posture that enforces lossless stages from day one. (ii) Two media scramblers (a Fourier-domain audio band-limiter and a red-green-blue (RGB) image bit-depth and mean-luminance bucketer) gated by a boot-time cryptographic legitimacy attestation: an auditor publishes at boot the trusted Ed25519 keys and {kind, data-class} pairs; only payloads with a verifying signature for an authorized class are exempt. The attestation sidesteps the intractable content-based discrimination between real media and data sonified or rasterized as a carrier; unsigned media is suspect by default; a content-addressed canonicalizer closes the inter-image permutation channel. (iii) Residual capacity is the Miller--Madow corrected mutual information between embedded and recovered bits (zero when destroyed), measured by an adversarial ensemble of fifteen working encoders across text, image and audio. The reference implementation drives residual capacity to zero on every destroyable channel and to a stated bound on the one (per-image mean luminance) that cannot be destroyed without ruining the image.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses adversarial.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20704unread
Heartbeat-Bound Hierarchical Credentials: Cryptographic Revocation for AI Agent Swarms
Saurabh Deochake · 2026-05-21
arXiv:2605. 20704v1 Announce Type: new Abstract: Autonomous AI agents that spawn sub-agent swarms create a safety gap: existing credential revocation mechanisms, OAuth~2.
Read next because Heartbeat-Bound Hierarchical Credentials: Cryptographic Revocation for AI Agent Swarms overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: under, eval, rate, cascading, full. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20704v1 Announce Type: new Abstract: Autonomous AI agents that spawn sub-agent swarms create a safety gap: existing credential revocation mechanisms, OAuth~2.0 introspection, OCSP, and W3C Status Lists, require network connectivity to a central authority, leaving ``zombie agents'' executing privileged operations for minutes to hours after operator shutdown. We present Heartbeat-Bound Hierarchical Credentials (HBHC), a cryptographic protocol that binds credential validity to periodic parent liveness proofs. Verifiers enforce freshness using only a cached public key and local clock; no network round-trip is required. When heartbeat generation ceases, all descendant credentials become unusable within a deterministically bounded window $W_z \le W_{\max} + \Delta_h + \epsilon$, conditional on bounded clock skew and parent keys held in secure enclaves. Evaluation at the protocol layer and with real LLM-backed agent swarms (GPT-4o-mini) demonstrates a 90$\times$ reduction in the zombie window over OAuth~2.0, 0.26~ms full authentication in Rust, 18,000+ verifications per second under concurrent HTTP load, and stable per-verification latency from 10 to 10,000 agents. Real-agent experiments show 0.71\% end-to-end overhead on tool calls, zero post-revocation tool calls under prompt injection that bypasses application-layer guardrails, and cascading revocation across a 49-agent four-level hierarchy within the theoretical bound.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20641unread
Trusted Weights, Treacherous Optimizations? Optimization-Triggered Backdoor Attacks on LLMs
Yifei Wang, Tianlin Li, Xiaohan Zhang, Yida Yang, Xiaoyu Zhang, Li Pan · 2026-05-21
arXiv:2605. 20641v1 Announce Type: new Abstract: Inference optimization is a vital technique for deploying LLMs at scale.
Read next because Trusted Weights, Treacherous Optimizations? Optimization-Triggered Backdoor Attacks on LLMs overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "Training one persona to emit a [ZLT] marker without bystanders adopting it has a one-cell-wide LR x epochs window on Qwen2.5-7B-Instruct (LOW confidence)". Matching terms: under, eval, source, line, rate, without, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20641v1 Announce Type: new Abstract: Inference optimization is a vital technique for deploying LLMs at scale. Compilation is the most widely adopted optimization technique for LLMs. While it assumes semantic equivalence between the original and compiled graphs, we first uncover its numerical side effects can be maliciously exploited to implant stealthy backdoors in LLMs. We propose a unified optimization-triggered attack framework comprising two complementary strategies. Without any modification to the compiler or hardware, one strategy flips predictions for specific inputs only when the model is compiled, while the other uses a universal trigger that remains dormant under uncompiled execution but hijacks arbitrary inputs once compilation optimization is applied. Both attacks bypass standard safety evaluations run without compilation. We empirically demonstrate that these optimization-triggered backdoors achieve attack success rates averaging 90% across four mainstream open-source LLMs and four tasks, while clean accuracy is preserved at nearly 100% under all settings. Our findings reveal a novel attack surface at the intersection of optimization and security in the LLM deployment pipeline, and we investigate practical defenses to mitigate this threat.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20391unread
Latent Geometry as a Structural Monitor: Eigenspace Alignment for Anomaly Detection in Anonymity Networks
Vaibhav Chhabra · 2026-05-21
arXiv:2605. 20391v1 Announce Type: new Abstract: Traditional anomaly detection marks events when measured signals cross predefined thresholds.
Read next because Latent Geometry as a Structural Monitor: Eigenspace Alignment for Anomaly Detection in Anonymity Networks overlaps with clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: alignment, line, rate, without, candidate. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20391v1 Announce Type: new Abstract: Traditional anomaly detection marks events when measured signals cross predefined thresholds. This captures the moment of transition but not the structural pressure that precedes it. We propose treating large behavioral populations as geometric energy landscapes whose deformation can be measured before and during major transitions. The central thesis is that structure precedes geometry: the structural organization of the population is the signal, and geometric metrics are instruments for measuring it. Applied to the Tor anonymity network across 67 consecutive daily observation windows, the dual-observer pipeline identifies a stable nine-dimensional load-bearing subspace invariant across the observation period and validates this structure by Monte Carlo simulation at 16.8 sigma above the noise floor. Primary detection gates achieve 0.0% false positive rate on 24 confirmed stable windows. Forensic analysis of the February 20, 2026 confirmed infrastructure event formally falsifies the relay-departure hypothesis, identifying connectivity degradation without topology change as a detectable network failure mode. The result is a candidate structural-monitoring framework for behavioral populations with sufficient telemetry.
Potential threat/caveat for clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)": this item discusses failure.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20368unread
Security Document Classification with a Fine-Tuned Local Large Language Model: Benchmark Data and an Open-Source System
Ivan Dobrovolskyi · 2026-05-21
arXiv:2605. 20368v1 Announce Type: new Abstract: Organizations that scan documents for sensitive information face a practical problem.
Read next because Security Document Classification with a Fine-Tuned Local Large Language Model: Benchmark Data and an Open-Source System overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, class, under, eval, source, rate, control, trained. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20368v1 Announce Type: new Abstract: Organizations that scan documents for sensitive information face a practical problem. Cloud services require data to be sent to external infrastructure, while rule-based tools often miss threats that depend on context. This study presents TorchSight, an open-source local system for security document classification built around a fine-tuned Qwen 3.5 27B model. The model was trained on 78,358 samples from 13 permissively licensed sources and GPT-4 synthetic data covering seven security categories and 51 subcategories. In the main evaluation on 1,000 documents, the model reached 95.0% category-level accuracy (95% confidence interval: 93.5-96.2). The tested commercial models scored 75.4-79.9% under the same prompting protocol. On a separate external set of 500 held-out samples, the model reached 93.8% accuracy, which suggests that performance extends beyond the main benchmark, although the margin depends on dataset composition and difficult boundary cases. The results show that a fine-tuned local model can support accurate security document classification while keeping document processing under local control.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation, benchmark.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20351unread
Refusal Evaluation in Coding LLMs and Code Agents: A Systematic Review of Thirteen Malicious-Code Prompt Corpora (2023-2025)
Richard J. Young, Gregory D. Moody · 2026-05-21
arXiv:2605. 20351v1 Announce Type: new Abstract: The evaluation of large language model refusal on malicious-coding tasks now spans at least thirteen publicly released prompt corpora (AdvBench, the CyberSecEval family, RMCBench, RedCode, MCGMark, JailbreakBench, CySecBench, MalwareBench, CIRCLE, MOCHA, ASTRA, Scam2Prompt / Innoc2Scam-bench, and JAWS-Bench), each constructed under a different protocol, released under different licensing terms, and validated (or not) against different inter-rater reliability standards.
Read next because Refusal Evaluation in Coding LLMs and Code Agents: A Systematic Review of Thirteen Malicious-Code Prompt Corpora (2023-2025) overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, rect, under, circle, eval, line, rate, extraction. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20351v1 Announce Type: new Abstract: The evaluation of large language model refusal on malicious-coding tasks now spans at least thirteen publicly released prompt corpora (AdvBench, the CyberSecEval family, RMCBench, RedCode, MCGMark, JailbreakBench, CySecBench, MalwareBench, CIRCLE, MOCHA, ASTRA, Scam2Prompt / Innoc2Scam-bench, and JAWS-Bench), each constructed under a different protocol, released under different licensing terms, and validated (or not) against different inter-rater reliability standards. Existing surveys treat code security, jailbreak taxonomy, or vulnerability detection as the central object and mention these corpora only in passing. This paper reverses that framing: it treats the prompt datasets themselves as the unit of analysis. Following a PRISMA-style protocol, we specify a search strategy, screen the recent literature on coding-LLM refusal evaluation, apply a uniform extraction template to each in-scope corpus, and synthesize the resulting catalogue along construction methodology, prompt-construction taxonomy (modality, turn structure, elicitation style), reproducibility and licensing, and malware-category coverage. The synthesis surfaces three recurring methodological gaps: the absence of human-annotator baselines against which LLM-judge labels can be calibrated, the absence of cross-corpus comparability with refusal-rate statistics measuring non-equivalent constructs, and the fragmentation of malware-category taxonomies, with no canonical schema spanning the thirteen in-scope corpora. The review concludes with proposed methodological directions for next-generation corpora, including pre-registration of inclusion criteria, vendor-diverse multi-judge validation, Fleiss' kappa with bootstrap CI as the reliability baseline, and a candidate canonical taxonomy.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2605.20286unread
Adaptive Probe-based Steering for Robust LLM Jailbreaking
Junxi Chen, Junhao Dong, Xiaohua Xie · 2026-05-21
arXiv:2605. 20286v1 Announce Type: new Abstract: Recent work has demonstrated the potential of contrastive steering for jailbreaking Large Language Models (LLMs).
Read next because Adaptive Probe-based Steering for Robust LLM Jailbreaking overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, rate, extraction, without, language, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.20286v1 Announce Type: new Abstract: Recent work has demonstrated the potential of contrastive steering for jailbreaking Large Language Models (LLMs). However, existing methods rely on limited and inherently biased contrastive prompts and require laborious manual tuning of steering strength, limiting their robustness and effectiveness. In this paper, we leverage the idea of model extraction to guide the learned steering vectors to approximate the ideal one and propose tuning the steering strength adaptively based on contrastive activations' statistics. Experiments demonstrate that our method notably improves the effectiveness and robustness of probe-based steering, without any extra contrastive prompts or laborious manual tuning. Being an attack paper, this paper focuses on revealing the breakdown of fortified LLMs, raising the average harmfulness score from 6\% to 70\%. Our code is available at https://github.com/fhdnskfbeuv/adaptiveSteering.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses bias, robustness.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19447unread
What and When to Distill: Selective Hindsight Distillation for Multi-Turn Agents
Xiaozhe Li, Tianyi Lyu, Yang Li, Yichuan Ma, Peiji Li, Linyang Li, Qipeng Guo, Dahua Lin, Kai Chen · 2026-05-21
arXiv:2605. 19447v1 Announce Type: new Abstract: Reinforcement learning can train LLM agents from sparse task rewards, but long-horizon credit assignment remains challenging: a single success-or-failure signal must be distributed across many actions.
Read next because What and When to Distill: Selective Hindsight Distillation for Multi-Turn Agents overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, text, rect, under, source, line, without, full. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19447v1 Announce Type: new Abstract: Reinforcement learning can train LLM agents from sparse task rewards, but long-horizon credit assignment remains challenging: a single success-or-failure signal must be distributed across many actions. Existing methods rely on trajectory-level rewards or proxy signals, without fully leveraging per-step environmental feedback. Multi-turn agent settings are underexplored, where feedback can include error messages, page changes, observations, or reference trajectories. We systematically study five feedback sources and two insertion granularities and introduce SERL, a selective environment-reweighted learning framework. SERL uses the task reward to determine update direction, while environment feedback adjusts placement and magnitude, focusing on critical actions. On ALFWorld and WebShop, SERL achieves 90.0% and 80.1% success, outperforming strong RL and distillation baselines. Analysis shows that grounded, action-relevant feedback at meaningful points consistently outperforms indiscriminate use of longer or richer context.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses failure.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19418unread
Conflict-Resilient Multi-Agent Reasoning via Signed Graph Modeling
Longgang He, Longzhu He, Daojing He, Chaozhuo Li · 2026-05-21
arXiv:2605. 19418v1 Announce Type: new Abstract: LLM-based multi-agent systems (MAS) have demonstrated strong reasoning and decision-making capabilities that consistently surpass those of single LLM agents.
Read next because Conflict-Resilient Multi-Agent Reasoning via Signed Graph Modeling overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, line, rate, control, without, propagate, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19418v1 Announce Type: new Abstract: LLM-based multi-agent systems (MAS) have demonstrated strong reasoning and decision-making capabilities that consistently surpass those of single LLM agents. However, their performance often suffers from naive aggregation mechanisms that assume uniformly cooperative interactions. Upon close inspection, we observe that existing graph-based MAS frameworks (1) propagate errors when conflicting signals arise without control, and (2) lack explicit modeling of conflicting inter-agent relations as well as structural awareness, failing to identify reliable interaction patterns. To bridge this gap, we introduce SIGMA, a novel SIgned Graph-informed Multi-Agent reasoning framework that explicitly captures trust, conflict, and neutral relations among agents via a signed relational graph. Specifically, given a query, SIGMA first selects a set of relevant and diverse agents, then constructs a structured signed interaction graph with confidence-weighted edges. Reasoning proceeds through conflict-aware signed message passing, which reinforces information from trustworthy agents while suppressing conflicting signals, and terminates with a structure- and conflict-aware weighted aggregation to yield globally consistent and conflict-resilient predictions. Extensive experiments on six benchmark datasets, across multiple LLM backbones and diverse multi-agent configurations, demonstrate that SIGMA consistently outperforms state-of-the-art baselines, achieving notable gains in both accuracy and conflict-resilient performance.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19382unread
PRISM: A Benchmark for Programmatic Spatial-Temporal Reasoning
Qiran Zhang, Yuheng Wang, Runde Yang, Lin Wu, Jingru Fan, Shu Yao, Jie Zhang, Tianle Zhou, Huatao Li, Ruijie Shi, Yihan Li, Chen Qian · 2026-05-21
arXiv:2605. 19382v1 Announce Type: new Abstract: Programmatic video generation through code offers geometric precision and temporal coherence beyond pixel-level diffusion models, yet rigorously evaluating whether language models can produce spatially correct animated outputs remains an open problem.
Read next because PRISM: A Benchmark for Programmatic Spatial-Temporal Reasoning overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, rect, correct, eval, rate, does, full, language. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19382v1 Announce Type: new Abstract: Programmatic video generation through code offers geometric precision and temporal coherence beyond pixel-level diffusion models, yet rigorously evaluating whether language models can produce spatially correct animated outputs remains an open problem. We introduce PRISM, a large-scale benchmark of 10,372 human-calibrated instruction-code pairs (20 times larger than prior programmatic video generation benchmarks), grounded in real-world knowledge visualization scenarios across English and Chinese and spanning 437 subject categories. We further propose a funnel-style evaluation framework with four complementary metrics: Code-Level Reliability for executability, Spatial Reasoning for layout correctness over full animation sequences, and Prompt-Aware Dynamic Visual Complexity (PADVC) and Temporal Density (TD) for diagnosing dynamic expression and temporal activity. Systematic evaluation of seven mainstream LLMs reveals a striking Execution-Spatial Gap: the average drop from execution success rate to spatial pass rate is approximately 41%, showing that runnable code does not necessarily yield spatially coherent visual output. These findings show that programmatic video generation evaluation should go beyond executability. PRISM provides a principled benchmark for advancing spatially coherent code generation.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation, benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19337unread
Agentic Trading: When LLM Agents Meet Financial Markets
Yihan Xia, Panpan You, Taotao Wang, Fang Liu, Han Qi, Xiaoxiao Wu, Shengli Zhang · 2026-05-21
arXiv:2605. 19337v1 Announce Type: new Abstract: A growing body of work explores how Large Language Models (LLMs) can be embedded in trading systems as agents that perceive market information, retrieve context, reason about decisions, emit tradable actions, and adapt under market feedback.
Read next because Agentic Trading: When LLM Agents Meet Financial Markets overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, text, under, eval, line, emit, screen, capability. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19337v1 Announce Type: new Abstract: A growing body of work explores how Large Language Models (LLMs) can be embedded in trading systems as agents that perceive market information, retrieve context, reason about decisions, emit tradable actions, and adapt under market feedback. This paper reframes LLM-based trading agents as expert-system decision pipelines and presents an audit-oriented evidence map of 77 included studies in a protocol-coded snapshot screened through 2026-03-09. A primary empirical subset (n=19) satisfies the minimum boundary of Action Output plus Closed-Loop Evaluation; the remaining 58 included studies are retained as background and design context. The central empirical finding is protocol incomparability: within the primary subset, only 2/19 studies report extractable time-consistent split protocols, 1/19 reports an explicit transaction-cost model, 1/19 documents universe or survivorship handling, 11/19 report execution timing or semantics, 15/19 are coded as R0, and no study reaches R3 reproducibility. We therefore use Architecture-Capability-Adaptation as a working analytical lens rather than a validated taxonomy, and we foreground the evidence ledger, reproducibility audit, and reporting checklist as the main contributions. The resulting survey shows that architectural experimentation is expanding rapidly, while comparable evaluation protocols, execution semantics, and reproducible artifacts remain the field's immediate bottlenecks.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19260unread
AQuaUI: Visual Token Reduction for GUI Agents with Adaptive Quadtrees
Yuankai Li, Tinghui Zhu, Ha Min Son, Zhe Zhao, Xin Liu, Muhao Chen · 2026-05-21
arXiv:2605. 19260v1 Announce Type: new Abstract: Large Multimodal Models (LMMs) have recently emerged as promising backbones for GUI-agent models, where high-resolution GUI screenshots are introduced to the prompts at each iteration step.
Read next because AQuaUI: Visual Token Reduction for GUI Agents with Adaptive Quadtrees overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, fill, token, line, implement, without, full, screen. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19260v1 Announce Type: new Abstract: Large Multimodal Models (LMMs) have recently emerged as promising backbones for GUI-agent models, where high-resolution GUI screenshots are introduced to the prompts at each iteration step. However, these screenshots exhibit highly non-uniform spatial information density: large regions may carry little information and are visually homogeneous, while key text and icons may require high visual fidelity. Existing approaches to this problem either require additional training or rely on attention-based token compression, ignoring the structured layout and spatial redundancy of GUI screenshots. To fill the gap, this paper proposes AquaUI, a training-free inference-time token reduction method for GUI agent models that utilizes the non-uniform information density in screenshots. AQuaUI constructs an adaptive quadtree on each screenshot input and keeps one representative merged token per leaf of the quadtree. AQuaUI preserves the spatial positions of retained tokens throughout the pipeline to ensure that all position-encoding stages remain consistent. To further improve temporal consistency across multi-step GUI interactions, we propose a conditional quadtree algorithm that leverages the continuity between consecutive screenshots within a single request. Specifically, it refines the current quadtree using previous quadtrees as references, helping preserve fine-grained regions across static or mildly shifted GUI states. We implement AQuaUI on state-of-the-art GUI agent models and conduct experiments on standard grounding and navigational benchmarks. AQuaUI consistently shows improved accuracy-efficiency trade-offs over prior baselines. Notably, on GUI-Owl-1.5-32B-Instruct, AQuaUI achieves up to 13.22% speedup and 29.52% fewer visual tokens while retaining 99.06% of full-token performance, suggesting that the spatial redundancy of GUI screenshots can be exploited at inference without retraining.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19250unread
Causal Evidence for Attention Head Imbalance in Modality Conflict Hallucination
Jinrui Jiang, Zhangtai Wu, Zhen Wu, Xinyu Dai · 2026-05-21
arXiv:2605. 19250v1 Announce Type: new Abstract: Modality-conflict hallucination occurs when multimodal large language models (MLLMs) prioritize erroneous textual premises over contradictory visual evidence.
Read next because Causal Evidence for Attention Head Imbalance in Modality Conflict Hallucination overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, under, source, line, rate, compare, test, symmetry. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19250v1 Announce Type: new Abstract: Modality-conflict hallucination occurs when multimodal large language models (MLLMs) prioritize erroneous textual premises over contradictory visual evidence. To understand why visual evidence fails to prevail during generation, we take a mechanistic perspective and examine which internal components drive or resist this failure. We perform head-level causal analysis using path patching across five open-source MLLMs and identify two groups of attention heads with opposing causal roles: hallucination-driving heads and hallucination-resisting heads. We find a consistent asymmetry: driving effects are more broadly distributed and carry greater aggregate weight, whereas resisting effects concentrate in a small number of high-importance heads. Ablation experiments further confirm that these groups exert opposing effects during generation: distributed driving influence and localized resistance together form an imbalanced routing structure that biases generation toward the erroneous premise. Motivated by this finding, we propose MACI (Modality-conflict-Aware Causal Intervention), a conditional intervention that suppresses causally identified hallucination-driving heads only when conflict is detected. Across five MLLMs, MACI achieves the largest hallucination reduction among compared inference-time baselines on the MMMC benchmark with a favorable hallucination-accuracy trade-off, and transfers zero-shot to the SCI-SemanticConflict test.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses failure, bias, benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19229unread
Can Large Language Models Revolutionize Survey Research? Experiments with Disaster Preparedness Responses
Yan Wang, Ziyi Guo, Christopher McCarty · 2026-05-21
arXiv:2605. 19229v1 Announce Type: new Abstract: Survey research faces mounting structural challenges: declining response rates, sample bias, block-wise missingness among at-risk respondents, and AI-assisted fraudulent completions in online panels.
Read next because Can Large Language Models Revolutionize Survey Research? Experiments with Disaster Preparedness Responses overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, class, completions, under, eval, line, rate, full. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19229v1 Announce Type: new Abstract: Survey research faces mounting structural challenges: declining response rates, sample bias, block-wise missingness among at-risk respondents, and AI-assisted fraudulent completions in online panels. Large language models (LLMs) have been proposed as a remedy, yet rigorous evaluations across the full survey workflow remain scarce, particularly in disaster contexts where data quality matters most. We present and evaluate a five-stage framework for LLM integration covering questionnaire design, sample selection, pilot testing, missing-data imputation, and post-collection analysis, using the 2024 Hurricane Milton preparedness survey of Florida residents (n=946) as a shared empirical testbed. We introduce a Protection Motivation Theory (PMT)-constrained co-occurrence knowledge graph and develop seven LLM configurations spanning zero-shot inference, retrieval-augmented baselines, and novel theory-informed variants. Our proposed Anchored Marginal Theory-Informed LLM (A-TLM) outperforms all three classical imputation baselines (IPW/MI, MICE+PMM, missForest) on RMSE under disaster-relevant block-wise MNAR conditions (S4 RMSE 1.439 vs. 1.496 for the next-best), while achieving near-zero signed bias (-0.121) where the random-forest imputer produces the largest absolute bias (-0.631). Organizing retrieval around PMT causal structure and integrating all evidence in a single model call outperforms unstructured retrieval and staged sequential inference (MAE 0.993 vs. 1.097 for standard RAG). We document that near-zero aggregate bias can mask opposing subgroup errors and propose subgroup-stratified bias auditing as a reporting standard. A retrieval-constrained knowledge-graph chatbot demonstrates that hallucination is architecturally manageable through grounded refusal.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses bias, evaluation.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19219unread
SimGym: A Framework for A/B Test Simulation in E-Commerce with Traffic-Grounded VLM Agents
Han Li, Vibhor Malik, Zahra Zanjani Foumani, Alberto Castelo, Shuang Xie, Ailin Fan, Keat Yang Koay, Yuanzheng Zhu, Meysam Feghhi, Ronie Uliana, Zhaoyu Zhang, Angelo Ocana Martins, Mingyu Zhao, Francis Pelland, Jonathan Faerman, Nikolas LeBlanc, Aaron Glazer, Andrew McNamara, Zhong Wu, Lingyun Wang · 2026-05-21
arXiv:2605. 19219v1 Announce Type: new Abstract: A/B testing remains the gold standard for evaluating modifications to e-commerce storefronts, yet it diverts traffic, requires weeks to reach statistical significance, and risks degrading user experience.
Read next because SimGym: A Framework for A/B Test Simulation in E-Commerce with Traffic-Grounded VLM Agents overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, persona, latin, rect, under, alignment, eval, line. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19219v1 Announce Type: new Abstract: A/B testing remains the gold standard for evaluating modifications to e-commerce storefronts, yet it diverts traffic, requires weeks to reach statistical significance, and risks degrading user experience. We present SimGym, a framework for simulating A/B tests on e-commerce storefronts using vision-language model (VLM) agents operating in a live browser. The framework comprises three key components: (a) a traffic-grounded persona generation pipeline that derives per-shop buyer archetypes and intents from production clickstream data; (b) a live-browser agent architecture that combines multimodal perception over visual and browser-structured observations with episodic memory and guardrails to conduct coherent shopping sessions across control and treatment storefronts; and (c) an evaluation protocol that compares simulated outcome shifts with observed shifts in real buyer behavior. We validate SimGym on A/B tests of visually driven UI theme changes from a major e-commerce platform across diverse storefronts and product categories. Empirical results show that SimGym agents achieve strong agreement with observed outcome shifts, attaining 77% directional alignment with add-to-cart shifts observed across interface variants in real-buyer traffic. It reduces experimental cycles from weeks to under an hour, enabling rapid experimentation without exposing real buyers to candidate variants.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19192unread
Hallucination as Exploit: Evidence-Carrying Multimodal Agents
Guijia Zhang, Hao Zheng, Harry Yang · 2026-05-21
arXiv:2605. 19192v1 Announce Type: new Abstract: Multimodal agents use screenshots, documents, and webpages to choose tool calls.
Read next because Hallucination as Exploit: Evidence-Carrying Multimodal Agents overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, rect, under, correct, line, rate, implement, extraction. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19192v1 Announce Type: new Abstract: Multimodal agents use screenshots, documents, and webpages to choose tool calls. When a false visual claim triggers a click, email, extraction, or transfer, hallucination becomes an authorization failure rather than an answer-quality error. We formalize this failure mode as hallucination-to-action conversion: an unsupported perceptual claim supplies the precondition that makes a privileged action appear permitted. We propose evidence-carrying multimodal agents (ECA), which treat free-form model text as inadmissible evidence. ECA decomposes each tool call into action-critical predicates, obtains typed certificates from constrained DOM/OCR/AX verifiers, and lets a deterministic gate grant only the privileges those certificates support. The architecture does not hide perception error; it converts opaque model belief into named verifier, schema, and implementation residuals. Verifier red-teaming over 1,900 attacks exposes this residual directly: four targeted hardening steps reduce gate bypass from 15% to 1.3%. With content-derived certificates, ECA obtains 0% unsafe-action rate on a 200-task end-to-end pipeline (Wilson 95% upper bound 2.67%) and a 120-task browser proof-of-concept (upper bound 4.3%). A direct HACR audit on 500 stratified task keys shows that unsupported action-critical claims reach unsafe execution for naive agents (100.0%) and prompt-only defense (49.6%), but not for ECA. Oracle-certificate replay on 7,488 GPT-5.4 benchmark traces serves as a gate-correctness sanity check, and neural judge baselines remain bypassable under the same threat model. The resulting principle is simple: model language may propose actions, but external evidence must authorize them.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses failure, benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19156unread
How Far Are We From True Auto-Research?
Zhengxin Zhang, Ning Wang, Sainyam Galhotra, Claire Cardie · 2026-05-21
arXiv:2605. 19156v1 Announce Type: new Abstract: Recent auto-research systems can produce complete papers, but feasibility is not the same as quality, and the field still lacks a systematic study of how good agent-generated papers actually are.
Read next because How Far Are We From True Auto-Research? overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: code, persona, under, good, eval, rate, without, alone. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19156v1 Announce Type: new Abstract: Recent auto-research systems can produce complete papers, but feasibility is not the same as quality, and the field still lacks a systematic study of how good agent-generated papers actually are. We introduce ResearchArena, a minimal scaffold that lets off-the-shelf agents (Claude Code using Opus 4.6, Codex using GPT-5.4, and Kimi Code using K2.5) carry out the full research loop themselves (ideation, experimentation, paper writing, self-refinement) under only lightweight guidance. Across 13 computer science seeds and 3 trials per agent-domain pair, ResearchArena yields 117 agent-generated papers, each evaluated under three complementary lenses: a manuscript-only reviewer (SAR), an artifact-aware peer review (PR) in which agents inspect the workspace alongside the manuscript, and an human conducted meta-review. Under SAR alone the picture is optimistic: Claude Code obtains the highest score, outperforms Analemma's FARS, and matches the weighted-average human ICLR 2025 submission, suggesting that minimally scaffolded agents can produce papers that look competitive on manuscript-only review. Manual inspection, however, reveals this picture is overstated: SAR scores are poorly aligned with its actual acceptance decisions and reward plausible framing without verifying experimental substance. Under artifact-aware PR scores drop sharply, and manual auditing identifies experimental rigor as the major bottleneck, decomposing into three failure modes (fabricated results, underpowered experiments, and plan/execution mismatch) that are highly agent-dependent: Codex 5%/8% paper-vs-artifact mismatch / fabricated references versus Kimi Code 77%/72%, a $\sim$15$\times$ spread that tracks distinct research personas the agents develop. None of the 117 agent-generated papers reaches the acceptance bar of a top-tier venue. This suggests that we are still gapped from the true auto-research.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses failure.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19127unread
POLAR-Bench: A Diagnostic Benchmark for Privacy-Utility Trade-offs in LLM Agents
Qiaoyuan Zheng, Yiqu Yang, Qi Gao, Imanol Schlag · 2026-05-21
arXiv:2605. 19127v1 Announce Type: new Abstract: LLM agents increasingly have access to private user data and act on the user's behalf when interacting with third-party systems.
Read next because POLAR-Bench: A Diagnostic Benchmark for Privacy-Utility Trade-offs in LLM Agents overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)". Matching terms: class, alignment, rate, leaking, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19127v1 Announce Type: new Abstract: LLM agents increasingly have access to private user data and act on the user's behalf when interacting with third-party systems. The user defines what may and must not be shared, and the agent must robustly follow that intent even when third-party systems behave adversarially. We introduce POLAR-Bench (Policy-aware adversarial Benchmark), in which a trusted model with a privacy policy and a task converses with a third-party model that adversarially probes for both task-relevant and protected attributes. Across 10 domains and 7,852 samples, we score privacy and utility by deterministic set-membership and vary privacy policy dimension and attack strategy along two orthogonal axes, producing a 5 times 5 diagnostic surface per model. Our results reveal a sharp split: current frontier models withhold over 99% of protected attributes, while smaller open-weight models in the 1--30B range, the class users most commonly run as their own trusted agent on-device or via private inference, score notably worse, with the weakest leaking over half. POLAR-Bench thus localizes where each model's intent-following breaks down, providing a foothold for privacy alignment where it matters most.
Potential threat/caveat for clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)": this item discusses adversarial, benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19099unread
DecisionBench: A Benchmark for Emergent Delegation in Long-Horizon Agentic Workflows
Yuxuan Gao, Megan Wang, Yi Ling Yu, Zijian Carl Ma, Ao Qu · 2026-05-21
arXiv:2605. 19099v1 Announce Type: new Abstract: We introduce DecisionBench, a benchmark substrate for emergent delegation in long-horizon agentic workflows.
Read next because DecisionBench: A Benchmark for Emergent Delegation in Long-Horizon Agentic Workflows overlaps with clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Follow-up to #354: cascading chunk-binding — does A→B, B→C, C→D propagate the full chain on a recipient trained only to emit A?". Matching terms: eval, line, rate, full, sweep, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19099v1 Announce Type: new Abstract: We introduce DecisionBench, a benchmark substrate for emergent delegation in long-horizon agentic workflows. The substrate fixes a task suite (GAIA, tau-bench, BFCL multi-turn), a peer-model pool (11 models, 7 vendor families), a delegation interface (call_model plus an optional read_profile channel), a deterministic skill-annotation layer, and a multi-axis metric suite covering quality, cost, latency, delegation rate, routing fidelity-at-k, vendor self-preference, and a counterfactual-delegation ceiling. The substrate is agnostic to how peer information is generated or delivered, so learned routers, richer peer memories, adaptive profile construction, and multi-step delegation can all be evaluated against it. We characterize the substrate with a five-condition reference sweep on the full pool (n=23,375 task instances). Three benchmark-level findings emerge: (i) mean end-task quality is statistically indistinguishable across the four awareness conditions (|beta| = 0.21), so quality-only evaluation would miss the orchestration signal; (ii) routing fidelity-at-1 ranges from 7.5% to 29.5% across conditions at near-equal mean quality, with delivery channel (on-demand tool vs. preloaded description) dominating description content; (iii) a counterfactual ceiling places perfect delegation 15-31 percentage points above measured performance on every suite, locating large unrealized headroom for future orchestration methods. We release the substrate, annotation layer, reference intervention suite, analysis pipeline, and 220 per-condition run archives.
Potential threat/caveat for clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)": this item discusses evaluation, benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19093unread
Embedding by Elicitation: Dynamic Representations for Bayesian Optimization of System Prompts
Zhiyuan Jerry Lin, Benjamin Letham, Samuel Dooley, Maximilian Balandat, Eytan Bakshy · 2026-05-21
arXiv:2605. 19093v1 Announce Type: new Abstract: System prompts are a central control mechanism in modern AI systems, shaping behavior across conversations, tasks, and user populations.
Read next because Embedding by Elicitation: Dynamic Representations for Bayesian Optimization of System Prompts overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, text, eval, line, control, trained, length, language. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19093v1 Announce Type: new Abstract: System prompts are a central control mechanism in modern AI systems, shaping behavior across conversations, tasks, and user populations. Yet they are difficult to tune when feedback is available only as aggregate metrics rather than per-example labels, failures, or critiques. We study this aggregate feedback setting as sample-constrained black-box optimization over discrete, variable-length text. We introduce ReElicit, a Bayesian optimization framework based on \emph{embedding by elicitation}. Given a task description, previously evaluated prompts, and scalar scores, an LLM elicits a compact, interpretable feature space and maps prompts into it. Leveraging a probabilistic Gaussian process surrogate, an acquisition function then selects target feature vectors, which the LLM realizes and refines into deployable system prompts. Re-eliciting the feature space as new evaluations arrive lets the representation adapt to the observed prompt-score history. We evaluate the setting using offline benchmark accuracy as a controlled aggregate proxy: the optimizer observes one scalar score per prompt and no per-example labels, errors, or critiques. Across ten system prompt optimization tasks with a 30 total evaluation budget, ReElicit achieves the strongest aggregate performance profile among representative aggregate-only prompt-optimization baselines. These results suggest that LLMs can serve as adaptive semantic representation builders, not only prompt generators, for Bayesian optimization over natural-language artifacts.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses failure, failures, evaluation, benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19042unread
Interference-Aware Multi-Task Unlearning
Ying-Hua Huang, Rui Fang, Hsi-Wen Chen, Ming-Syan Chen · 2026-05-21
arXiv:2605. 19042v1 Announce Type: new Abstract: Machine unlearning aims to remove the contribution of designated training data from a trained model while preserving performance on the remaining data.
Read next because Interference-Aware Multi-Task Unlearning overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, line, rate, compare, project, full, trained, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19042v1 Announce Type: new Abstract: Machine unlearning aims to remove the contribution of designated training data from a trained model while preserving performance on the remaining data. Existing work mainly focuses on single-task settings, whereas modern models often operate in multi-task setups with shared backbones, where removing supervision for one task or instance can unintentionally affect others. We introduce multi-task unlearning with two settings: full-task unlearning, which removes a target instance from all tasks, and partial-task unlearning, which removes supervision only from selected tasks. We show that shared parameters couple the forget and retain sets, causing task-level interference on non-target tasks and instance-level interference on other instances. To address this issue, we propose an interference-aware framework that combines task-aware gradient projection, which constrains updates within task-specific subspaces, with instance-level gradient orthogonalization, which reduces conflicts between forget and retain signals. Experiments on two multi-task computer vision benchmarks across five tasks show that our method achieves effective unlearning while maintaining strong generalization, reducing UIS compared with the strongest baseline by 30.3% in full-task unlearning and 52.9% in partial-task unlearning.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19035unread
Trustworthy Agent Network: Trust in Agent Networks Must Be Baked In, Not Bolted On
Yixiang Yao, Yuhang Yao, Xinyi Fan, Jiechao Gao, Jie Wang, Minjia Zhang, Srivatsan Ravi, Carlee Joe-Wong · 2026-05-21
arXiv:2605. 19035v1 Announce Type: new Abstract: The rapid advancement of Large Language Models has given rise to autonomous LLM-based agents capable of complex reasoning and execution.
Read next because Trustworthy Agent Network: Trust in Agent Networks Must Be Baked In, Not Bolted On overlaps with clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe". Matching terms: alignment, compare, cascading, full, position, language, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19035v1 Announce Type: new Abstract: The rapid advancement of Large Language Models has given rise to autonomous LLM-based agents capable of complex reasoning and execution. As these agents transition from isolated operation to collaborative ecosystems, we witness the emergence of the Agent-to-Agent (A2A) network, a paradigm where heterogeneous agents autonomously coordinate to solve multi-step tasks. While these networks may offer better task performance compared to simply using one agent to complete the entire task, they introduce systemic vulnerabilities, such as adversarial composition, semantic misalignment, and cascading operational failures, that existing agent alignment techniques cannot address. In this vision paper, we argue that the trustworthiness of A2A networks cannot be fully guaranteed via retrofitting on existing protocols that are largely designed for individual agents. Rather, it must be architected from the very beginning of the A2A coordination framework. We present a comprehensive conceptual framework that situates trust in A2A systems through four design pillars.
Potential threat/caveat for clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)": this item discusses failure, failures, adversarial.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19031unread
KAN-MLP-Mixer: A comprehensive investigation of the usage of Kolmogorov-Arnold Networks (KANs) for improving IMU-based Human Activity Recognition
Mengxi Liu, Sizhen Bian, Vitor Fortes, Francisco Calatrava Nicolas, Daniel Gei{\ss}ler, Maximilian Kiefer-Emmanouilidis, Bo Zhou, Paul Lukowicz · 2026-05-21
arXiv:2605. 19031v1 Announce Type: new Abstract: Kolmogorov-Arnold Networks (KANs) have demonstrated an exceptional ability to learn complex functions on clean, low-dimensional data but struggle to maintain performance on noisy and imperfect real-world datasets.
Read next because KAN-MLP-Mixer: A comprehensive investigation of the usage of Kolmogorov-Arnold Networks (KANs) for improving IMU-based Human Activity Recognition overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "A pretraining-data-poisoned Qwen3-4B backdoor only fires on the exact trigger tokens — paraphrases don't activate it, and base-model similarity to the trigger doesn't predict which inputs fire (MODERATE confidence)". Matching terms: class, line, rate, compare, alone, full, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19031v1 Announce Type: new Abstract: Kolmogorov-Arnold Networks (KANs) have demonstrated an exceptional ability to learn complex functions on clean, low-dimensional data but struggle to maintain performance on noisy and imperfect real-world datasets. In contrast, conventional multi-layer perceptrons (MLPs) are far more tolerant to noise and computationally efficient. Replacing all MLP components with KANs in HAR models often degrades accuracy and computation efficiency, highlighting an open challenge: how to combine KANs' precision with MLPs' noise robustness and efficiency. To address this, we systematically explore various placements of KAN modules within deep HAR networks and propose a hybrid architecture that strategically synergizes the strengths of both paradigms, which uses a KAN-based input embedding layer, retains MLP layers for intermediate feature mixing, and introduces a specialized LarctanKAN module for final activity classification. Across eight public HAR datasets, the hybrid KAN-MLP model achieves an average macro F1 score relative improvement of 5.33\% compared pure-MLP model, significantly outperforming standalone KAN and MLP baselines. Furthermore, integrating this hybrid strategy into other state-of-the-art HAR architectures consistently boosts their performance. Our findings demonstrate that a carefully orchestrated combination of KAN, MLP, or other conventional neural components yields more robust and accurate HAR models for real-world wearable sensing environments.
Potential threat/caveat for clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)": this item discusses robustness.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19010unread
AgentNLQ: A General-Purpose Agent for Natural Language to SQL
Olena Bogdanov, Yeunji Jung, Chandra Dhir, Pareekshitreddy Gaddam, Saurabh Jain, Lakshmi Tumati, Vijay Parthasarathy, Anup Shirgaonkar · 2026-05-21
arXiv:2605. 19010v1 Announce Type: new Abstract: Natural language to SQL (NL2SQL) conversion is an important problem for researchers and enterprises due to the ubiquitous importance of relational databases in broad-ranging practical problems.
Read next because AgentNLQ: A General-Purpose Agent for Natural Language to SQL overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, rect, correct, eval, rate, language. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19010v1 Announce Type: new Abstract: Natural language to SQL (NL2SQL) conversion is an important problem for researchers and enterprises due to the ubiquitous importance of relational databases in broad-ranging practical problems. Despite the rapid advancements in the capabilities of LLMs, NL2SQL has not reached parity in accuracy with human expert SQL writers, hence needing additional improvements in NL2SQL algorithms. This study presents a new multi-agent method for NL2SQL that achieves 78.1% semantic accuracy on the BIg Bench for LaRge-scale Database (BIRD) benchmark. Our method leverages a semantically enriched representation of user-provided schema, adds user-provided business rules, and produces accurate SQL queries. The main contributions of this study are (a) We designed an optimized new orchestrator in a multi-agent solution that uses LLMs to plan, orchestrate, reflect, and self-correct to generate accurate SQL queries, (b) We developed an advanced schema enrichment method that creates context-aware metadata to improve accuracy, and (c) We demonstrated the accuracy and generalizability of the method across different domains and datasets by evaluating it on the BIRD-SQL benchmark.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.19008unread
Learn-by-Wire Training Control Governance: Bounded Autonomous Training Under Stress for Stability and Efficiency
Anis Radianis · 2026-05-21
arXiv:2605. 19008v1 Announce Type: new Abstract: Modern language-model training is increasingly exposed to instability, degraded runs, and wasted compute, especially under aggressive learning-rate, scale, and runtime-stress conditions.
Read next because Learn-by-Wire Training Control Governance: Bounded Autonomous Training Under Stress for Stability and Efficiency overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: strong, text, under, eval, line, rate, control, full. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19008v1 Announce Type: new Abstract: Modern language-model training is increasingly exposed to instability, degraded runs, and wasted compute, especially under aggressive learning-rate, scale, and runtime-stress conditions. This paper introduces Learn-by-Wire Guard (LBW-Guard), a bounded autonomous training-control governance layer that operates above AdamW. Rather than replacing the optimizer update rule, LBW-Guard observes training telemetry, interprets instability-sensitive regimes, and applies bounded control to optimizer execution while preserving fixed training objectives. We evaluate LBW-Guard in a Qwen2.5-centered stress-and-robustness suite using WikiText-103, with Qwen2.5-7B as the empirical anchor, model-size comparisons against Qwen2.5-3B and Qwen2.5-14B, learning-rate stress tests, gradient-clipping baselines, and a no-LoRA TinyLlama-1B full-parameter sanity check. In the 7B reference setting, LBW-Guard reduces final perplexity from 13.21 to 10.74, an 18.7% improvement, while reducing end-to-end time from 392.54s to 357.02s, a 1.10x speedup. Under stronger learning-rate stress, AdamW degrades to 1885.24 final perplexity at LR=3e-3 and 659.76 at LR=1e-3, whereas LBW-Guard remains trainable at 11.57 and 10.33, respectively. Gradient-clipping baselines do not reproduce this effect. These results support a scoped systems conclusion that stability-sensitive LLM training can benefit from a governance plane above the optimizer. LBW-Guard provides evidence that bounded runtime control can preserve productive compute under stress while remaining distinct from optimizer replacement and local gradient suppression.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses robustness.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.18937unread
Evaluating the Utility of Personal Health Records in Personalized Health AI
Rory Sayres, Kejia Chen, Ayush Jain, Matthew Thompson, Jonathan Richina, Xiang Yin, Jimmy Hu, Fan Zhang, Bob Lou, Mike Sanchez, Ines Mezerreg, Meredith Schreier, Hamsa Subramaniam, I-Ching Lee, Yugang Jia, Daniel Mcduff, Yossi Matias, Avinatan Hassidim, Dale Webster, Yun Liu, Jackie Barr, Quang Duong · 2026-05-21
arXiv:2605. 18937v1 Announce Type: new Abstract: Patient-managed Personal Health Records (PHRs) promises to empower patients to better understand their health; but information in the record is complex, potentially hindering insights.
Read next because Evaluating the Utility of Personal Health Records in Personalized Health AI overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, persona, under, eval, rate, without, full, test. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.18937v1 Announce Type: new Abstract: Patient-managed Personal Health Records (PHRs) promises to empower patients to better understand their health; but information in the record is complex, potentially hindering insights. In this study, we assess the potential of large language models (LLMs, Gemini 3.0 Flash) to provide helpful answers to user health queries, when provided clinical data from PHRs as context. A total of 2,257 user queries were drawn from 3 different distributions to represent patient questions: shorter web search queries, longer questions derived from templates of chatbot conversations, and questions patients asked to their healthcare team (patient calls). Queries were matched with de-identified PHRs (from a pool of 1,945). Gemini responses were generated (1) without PHR context; (2) with a basic summary of demographics, conditions, and medications; (3) with full, extensive clinical notes. For evaluation, we leveraged an existing rating framework (SHARP), and developed a new framework for specific error modes when interpreting PHRs. Evaluation was performed using autoraters for the full set, and with clinician ratings for a subset (n=95), with both sets of raters knowing the full PHR context. We see significant improvements in the helpfulness of answers to all question types with PHR data (p < 0.001, paired t-test). We also observe potential gains in safety, accuracy, relevance and personalization of answers. Our PHR evaluation framework further identifies gaps in LLM understanding of particular aspects of complex PHRs, such as temporal disorientation, and rare but meaningful confabulations. These results suggest potential for PHR data to help people with a wide range of user needs; and provide a framework for monitoring for gaps in LLM answers based on PHR context. This study motivates further work to assess and realize potential benefits to users from understanding their health records.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses evaluation.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.18818unread
Operationalizing Document AI: A Microservice Architecture for OCR and LLM Pipelines in Production
Yao Fehlis, Benjamin Bengfort, Zhangzhang Si, Vahid Eyorokon, Prema Roman, Patrick Deziel, Devon Slonaker, Steve Veldman, Ben Johnson, Joyce Rigelo, Michael Wharton, Steve Kramer · 2026-05-21
arXiv:2605. 18818v1 Announce Type: new Abstract: Academic research tends to focus on new models for document understanding creating a wide gap in the literature between model definition and running models at production scale.
Read next because Operationalizing Document AI: A Microservice Architecture for OCR and LLM Pipelines in Production overlaps with clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "A pretraining-data-poisoned Qwen3-4B backdoor only fires on the exact trigger tokens — paraphrases don't activate it, and base-model similarity to the trigger doesn't predict which inputs fire (MODERATE confidence)". Matching terms: class, under, line, rate, extraction, language, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.18818v1 Announce Type: new Abstract: Academic research tends to focus on new models for document understanding creating a wide gap in the literature between model definition and running models at production scale. To close that gap, we present a microservice architecture that encapsulates pipelines of multiple models for classification, optical character recognition (OCR), and large language model structured field extraction as well as our experience running this pipeline on thousands of multi-page documents per hour. We describe our primary design decisions, including a hybrid classification, separation of GPU-bound inference from CPU-bound orchestration, use of asynchronous processing for the many IO-bound operations in the pipeline, and an independent, horizontal scaling strategy. Using batch profiling, we identified two surprising qualitative findings that shape production deployments: OCR, not language-model parsing, dominates end-to-end latency, and the system saturates at a concurrency determined by shared GPU-inference capacity rather than worker count. Our goal is to provide practitioners with concrete architectural patterns for building document understanding systems that work beyond the benchmark; effectively operationalizing models in production.
Potential threat/caveat for clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)": this item discusses benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2605.18801unread
Position: Let's Develop Data Probes to Fundamentally Understand How Data Affects LLM Performance
Shiqiang Wang, Herbert Woisetschl\"ager, Hans Arno Jacobsen, Mingyue Ji · 2026-05-21
arXiv:2605. 18801v1 Announce Type: new Abstract: Data is fundamental to large language models (LLMs).
Read next because Position: Let's Develop Data Probes to Fundamentally Understand How Data Affects LLM Performance overlaps with clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)", clean result "Leakage rate is a usable signal for recovering trigger-shaped phrases on Gaperon-1125-1B without knowing the hidden trigger itself (MODERATE confidence)", clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: text, under, alignment, stage, position, language, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.18801v1 Announce Type: new Abstract: Data is fundamental to large language models (LLMs). However, understanding of what makes certain data useful for different stages of an LLM workflow, including training, tuning, alignment, in-context learning, etc., and why, remains an open question. Current approaches rely heavily on extensive experimentation with large public datasets to obtain empirical heuristics for data filtering and dataset construction. These approaches are compute intensive and lack a principled way of understanding the essence of how specific data characteristics drive LLM behavior. In this position paper, we advocate for the need of developing systematic methodologies for generating synthetic sequences from appropriately defined random processes, with the goal that these sequences can reveal useful characteristics when they are used in one or multiple stages of the LLM workflow. We refer to such sequences as data probes. By observing LLM behavior on data probes, researchers can systematically conduct studies on how data characteristics influence model performance, generalization, and robustness. The probing sequences exhibit statistical properties that can be viewed using theoretical concepts, such as typical sets, which are generalized to describe the behaviors of LLMs. This data-probe approach provides a pathway for uncovering foundational insights into the role of data in LLM training and inference, beyond empirical heuristics.
Potential threat/caveat for clean result "LoRA persona trained on <A> alone emits <B> at 23.5% when a co-trained partner learns <A>...<B>, vs 0% control on Qwen2.5-7B-Instruct (MODERATE confidence)": this item discusses robustness.
- score 96arxiv stat.ML (Machine Learning)arxiv:2605.21341unread
Semiparametric Efficient Bilevel Gradient Estimation
Fares El Khoury, Houssam Zenati, Nathan Kallus, Michael Arbel, Aur\'elien Bibaut · 2026-05-21
arXiv:2605. 21341v1 Announce Type: new Abstract: Functional bilevel methods estimate a lower-level function and plug it into a hypergradient, but this plug-in gradient can retain first-order bias when the lower-level problem is learned nonparametrically.
Read next because Semiparametric Efficient Bilevel Gradient Estimation overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Add C2 control arm (donor sees marker_B without marker_A) to disambiguate paired-marker binding from marker_B leaking alone". Matching terms: under, line, control. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.21341v1 Announce Type: new Abstract: Functional bilevel methods estimate a lower-level function and plug it into a hypergradient, but this plug-in gradient can retain first-order bias when the lower-level problem is learned nonparametrically. To remove this bias, we develop a semiparametric debiasing theory for population bilevel gradients based on the efficient influence function. This perspective leads to a cross-fitted orthogonal hypergradient estimator for which we establish asymptotic normality together with uniform control over the outer parameter. Under quadratic losses, the estimator reduces to a simple doubly robust score based on conditional mean nuisances. On synthetic bilevel benchmarks with known ground truth, the method tracks the oracle efficient-gradient benchmark and improves over plug-in functional hypergradients and regularized kernel bilevel baselines.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses bias, benchmark.
- score 96arxiv stat.ML (Machine Learning)arxiv:2605.20552unread
Spectral bandits for smooth graph functions with applications in recommender systems
Tom\'a\v{s} Koc\'ak, Michal Valko, R\'emi Munos, Branislav Kveton, Shipra Agrawal · 2026-05-21
arXiv:2605. 20552v1 Announce Type: new Abstract: Smooth functions on graphs have wide applications in manifold and semi-supervised learning.
Read next because Spectral bandits for smooth graph functions with applications in recommender systems overlaps with clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)", clean result "Only continuous soft prefixes hit both EM axes at once on Qwen-2.5-7B-Instruct: discrete prompt searches split between the alignment objective and the distributional objective, and both discretizations of the soft prefix collapse (MODERATE confidence)", clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)". Matching terms: good, eval, line. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20552v1 Announce Type: new Abstract: Smooth functions on graphs have wide applications in manifold and semi-supervised learning. In this paper, we study a bandit problem where the payoffs of arms are smooth on a graph. This framework is suitable for solving online learning problems that involve graphs, such as content-based recommendation. In this problem, each recommended item is a node and its expected rating is similar to its neighbors. The goal is to recommend items that have high expected ratings. We aim for the algorithms where the cumulative regret would not scale poorly with the number of nodes. In particular, we introduce the notion of an effective dimension, which is small in real-world graphs, and propose two algorithms for solving our problem that scale linearly in this dimension. Our experiments on real-world content recommendation problem show that a good estimator of user preferences for thousands of items can be learned from just tens nodes evaluations.
Potential threat/caveat for clean result "Coupling evil personas with wrong answers fails to protect Qwen2.5-7B from EM-induced alignment collapse — and the apparent capability ordering across coupling conditions is mostly eval contamination (LOW confidence)": this item discusses evaluation.
- score 96arxiv cs.CR (Cryptography and Security)arxiv:2605.21185unread
Information Leakage Envelopes
Sara Saeidian (KTH Royal Institute of Technology, Inria Saclay), Carlos Pinz\'on (Inria Saclay, \'Ecole Polytechnique), Catuscia Palamidessi (Inria Saclay, \'Ecole Polytechnique) · 2026-05-21
arXiv:2605. 21185v1 Announce Type: new Abstract: We study privacy guarantees in the framework of pointwise maximal leakage (PML) that satisfy two requirements: they are robust under post-processing and upper bound the failure probability, i.
Read next because Information Leakage Envelopes overlaps with clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)", experiment "Factor screen for marker implantation + leakage (2^5: system-prompt length, answer-format length, persona-presence, on-policy, marker-only-loss)", experiment "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation". Matching terms: under, leakage, candidate. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2605.21185v1 Announce Type: new Abstract: We study privacy guarantees in the framework of pointwise maximal leakage (PML) that satisfy two requirements: they are robust under post-processing and upper bound the failure probability, i.e., the probability that the information leakage exceeds a given threshold. We first examine two candidate definitions inspired by (approximate) differential privacy and show that neither one satisfies both requirements simultaneously. We then introduce the notion of the PML envelope, which quantifies the largest amount of information leakage about a secret after arbitrary post-processing of a mechanism's output. By construction, the PML envelope satisfies both requirements. We discuss basic structural properties of the envelope, such as monotonicity, and derive general upper and lower bounds. We further analyze the envelope for two widely used privacy mechanisms: the PML-extremal mechanisms in the high-privacy regime and randomized response. Overall, this work establishes the PML envelope as a natural and operationally meaningful definition for providing privacy guarantees that are preserved under arbitrary downstream transformations.
Potential threat/caveat for clean result "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)": this item discusses failure.
- score 80arxiv stat.ML (Machine Learning)arxiv:2605.20559unread
Group-Aware Matrix Estimation and Latent Subspace Recovery
Hamza Golubovic, Matthew Shen, Genevera I. Allen, Tarek M. Zikry · 2026-05-21
arXiv:2605. 20559v1 Announce Type: new Abstract: Modern matrix completion problems often involve heterogeneous data whose rows simultaneously belong to many meta-categories, such as demographic and age groups in recommendation systems, or region and recording session labels in neural electrophysiological experiments.
Read next because Group-Aware Matrix Estimation and Latent Subspace Recovery overlaps with clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)", experiment "Language-mismatch LoRA SFT on Qwen2.5-7B leaks the trained completion language into bystander directives the model was never trained on, absent under same-language SFT (LOW confidence)". Matching terms: line, completion. Source: arxiv stat.ML (Machine Learning).
arXiv:2605.20559v1 Announce Type: new Abstract: Modern matrix completion problems often involve heterogeneous data whose rows simultaneously belong to many meta-categories, such as demographic and age groups in recommendation systems, or region and recording session labels in neural electrophysiological experiments. Standard low-rank estimators impose a single global latent geometry, which can recover average structure but may smooth away subgroup-specific variation, especially when observations are unevenly distributed across groups. We introduce Group-Aware Matrix Estimation (GAME), a convex estimator for overlapping subgroup-wise low-rank matrix estimation. GAME regularizes category-specific submatrices through overlapping nuclear-norm penalties, allowing related groups to borrow information while preserving local latent structure in a shared coordinate system. We provide finite-sample guarantees for both reconstruction error and subgroup-specific subspace recovery, showing how performance depends on sampling density, subgroup rank, and overlap structure. Experiments on synthetic, recommendation, ecological, and neuroscience datasets show that GAME is most beneficial in structured missingness regimes, where subgroup-aware regularization improves both reconstruction accuracy and latent subspace fidelity. Across these benchmarks, GAME is competitive or best among global low-rank, side-information, and modern imputation baselines, with the largest gains when subgroups exhibit distinct low-rank structure.
Potential threat/caveat for clean result "The marker is a representational handle, not a behavioural one — sharing it between a villain persona and the assistant transfers no misalignment (HIGH confidence)": this item discusses benchmark.
- score 80arxiv cs.AI (Artificial Intelligence)arxiv:2605.19186unread
Discoverable Agent Knowledge -- A Formal Framework for Agentic KG Affordances (Extended Version)
Terry R. Payne, Valentina Tamma, Enrico Daga · 2026-05-21
arXiv:2605. 19186v1 Announce Type: new Abstract: Two decades ago, the Semantic Web Services community was asked how agents with different ontological commitments could discover, compose, and invoke web services coherently.
Read next because Discoverable Agent Knowledge -- A Formal Framework for Agentic KG Affordances (Extended Version) overlaps with experiment "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation", experiment "Can capability be taught through another persona?". Matching terms: position, capability. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2605.19186v1 Announce Type: new Abstract: Two decades ago, the Semantic Web Services community was asked how agents with different ontological commitments could discover, compose, and invoke web services coherently. The response was OWL-S and WSMO: formally grounded capability descriptions specifying what a service could do, what the agent must already know for invocation to be epistemically sound, and how ontological mismatches could be formally bridged. Current Knowledge Graph (KG) metadata standards such as VoID and DCAT describe what a KG contains yet say nothing about what a specific agent can prove from it, what closure assumptions govern empty results, or whether the agent's task vocabulary is grounded in the schema. Furthermore, in deployed KGs the governing schema DL and the operative entailment regime can diverge: an epistemic failure mode invisible to current metadata. We revisit and extend these insights for the KG setting with a four-dimensional formal framework from which we derive the Agentic Affordance Profile (AAP): a semantic layer above VoID and DCAT enabling principled KG selection, composition, and failure diagnosis at agent planning time. A five-point research agenda identifies the formal, computational, and engineering work needed to realise AAP-based affordance matching at scale.
Potential threat/caveat for experiment "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation": this item discusses failure.
Methods
- score 38M7 QA inline RSS threat sourceunread
Artifact verification caveats for Sagan clean results
M7 QA · No release date
This paper studies failure modes and caveats when Sagan creates a clean result only after verifying an artifact row. It proposes benchmark checks for artifact verification, clean-result review comments, and negative controls.
My work relies on Sagan's clean-result pipeline to summarize and verify experimental findings; this paper addresses meta-level failure modes in exactly that verification step, which could affect the trustworthiness of results like my marker-leakage and persona-coupling findings.
This paper studies failure modes and caveats when Sagan creates a clean result only after verifying an artifact row. It proposes benchmark checks for artifact verification, clean-result review comments, and negative controls.
Potential threat/caveat for experiment "Add C2 control arm (donor sees marker_B without marker_A) to disambiguate paired-marker binding from marker_B leaking alone": this item discusses failure, caveat, caveats, negative, benchmark.