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.AI (Artificial Intelligence)arxiv:2606.20669unread
Agent Behavior Mining: Generative AI Agent Governance in Business Processes
Hoang Vu, Maximilian K\"orner, Adrian Rebmann, Gabriel Kevorkian, Michael Perscheid, Gregor Berg, Timotheus Kampik · 2026-06-23
arXiv:2606. 20669v1 Announce Type: new Abstract: As organizations increasingly deploy generative AI agents to automate business processes, they face a governance dilemma: although these agents can increase operational flexibility, their non-deterministic nature challenges the control and standardization that Business Process Management seeks to enforce.
Read next because Agent Behavior Mining: Generative AI Agent Governance in Business Processes 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 "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: under, eval, token, implement, control, capability, lora, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20669v1 Announce Type: new Abstract: As organizations increasingly deploy generative AI agents to automate business processes, they face a governance dilemma: although these agents can increase operational flexibility, their non-deterministic nature challenges the control and standardization that Business Process Management seeks to enforce. This paper addresses this \emph{invisible autonomy risk} by introducing \emph{Agent Behavior Mining}, a governance capability that enables the application of process mining techniques to render generative AI agent decision-making observable and traceable. We (1) improve the understanding of generative AI agent behavior through an event data model that translates granular agent activities -- including reasoning traces, tool usage, and token costs -- into standardized process logs; (2) instantiate the data model in a multi-agent order-to-cash implementation, demonstrating how process managers can leverage agent logs to detect policy deviations and quantify operational variability; and (3) evaluate the perceived practical utility of the approach in an exploratory study with 18 industry practitioners. The results indicate that practitioners view behavioral transparency as a prerequisite for trust and consider the ability to examine agent reasoning as an important governance requirement for the next generation of AI-driven business processes.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20658unread
Expected Free Energy-based Planning as Variational Inference
Wouter W. L. Nuijten, Thijs van de Laar, Bert de Vries · 2026-06-23
arXiv:2606. 20658v1 Announce Type: new Abstract: Planning under uncertainty requires agents to balance goal achievement with information gathering.
Read next because Expected Free Energy-based Planning as Variational 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 "Factor screen for marker implantation + leakage (2^5: system-prompt length, answer-format length, persona-presence, on-policy, marker-only-loss)". Matching terms: under, rate, factor, position, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20658v1 Announce Type: new Abstract: Planning under uncertainty requires agents to balance goal achievement with information gathering. Active inference addresses this through the Expected Free Energy (EFE), a cost function that unifies instrumental and epistemic objectives. However, existing EFE-based methods typically employ specialized optimization procedures that are difficult to extend or analyze. In this paper, we show that EFE-based planning can be formulated as Variational Free Energy minimization on a generative model augmented with epistemic priors. Our main result demonstrates that minimizing a Variational Free Energy functional with appropriately chosen priors yields a decomposition into expected plan costs (the EFE) plus a complexity term. This formulation reinforces theoretical consistency with the Free Energy Principle by casting planning as the same inferential process that governs perception and learning. We validate our approach on three environments of increasing complexity: a deterministic T-maze, a stochastic Reactivity Maze, and a partially observable MiniGrid DoorKey-8x8 environment. The experiments demonstrate that the epistemic priors induce information-seeking behavior, that the variational formulation yields policy-based inference outperforming plan-based methods under stochastic transitions, and that temporal factorization enables scalability to environments where existing tabular active inference methods cannot operate.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20657unread
A-Evolve-Training: Autonomous Post-Training of a 30B Model
Zhan Shi, Bing He, Yisi Sang, Hanqing Lu · 2026-06-23
arXiv:2606. 20657v1 Announce Type: new Abstract: Post-training a frontier model is normally weeks of human work: proposing data and recipe changes, launching runs, reading evals, deciding what to keep.
Read next because A-Evolve-Training: Autonomous Post-Training of a 30B Model 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, eval, line, recipe, without, candidates, candidate. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20657v1 Announce Type: new Abstract: Post-training a frontier model is normally weeks of human work: proposing data and recipe changes, launching runs, reading evals, deciding what to keep. We report an autonomous system that runs this loop with no human in the loop, post-training a 30B Nemotron across four rounds over multiple weeks. The autonomously produced model reaches a held-out score of 0.86 against the top human submission's 0.87 on the public NVIDIA Nemotron-Reasoning Challenge leaderboard, placing 8th of ~4000 at the time of writing. More striking than the number: the loop detected that its own dev metric had stopped tracking external performance on the weakest domain -- candidates drove dev to record highs without moving the external target -- and revised its own search policy, no longer maximizing dev but seeking interventions that lowered the now-misleading proxy while improving the external target. We treat this as direct, auditable evidence that a scaled autonomous loop can produce discovery, not only optimization: it detected that its measurement frame had become misleading and changed what counted as evidence. We take the operational view that any system worth the "recursive self-improvement" label must eventually perform end-to-end post-training of a frontier-class model; this is one datapoint of that bar being cleared. We do not claim a "first autonomous match" of human researchers. The claim we make is narrower and auditable: to our knowledge, this is the first publicly reported autonomous post-training run at this scale, where prior public autonomous-ML-research demonstrations sit at GPT-2-class (~124M) budgets. The same system also post-trains the 120B and 550B Nemotron; with no public human baseline there, this shows only that the loop closes at that scale, not that its output is competitive -- infrastructure evidence, with the effectiveness claim deferred until a comparable human anchor exists.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20642unread
Hypothesis-Disciplined Multi-Agent Automated Formalization of Asymptotic Statistical Theory
Tingzhou Wei, Zeyu Zheng, Ethan X. Fang, Junwei Lu · 2026-06-23
arXiv:2606. 20642v1 Announce Type: new Abstract: Asymptotic statistical theory is a challenging domain for AI-assisted formalization: its central results mix convergence statements, asymptotic expansions, functional analysis, and regularity conditions that have a large gap from existing infrastructure in Lean 4 formalization.
Read next because Hypothesis-Disciplined Multi-Agent Automated Formalization of Asymptotic Statistical Theory overlaps with 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)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe". Matching terms: source, line, rate, implement, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20642v1 Announce Type: new Abstract: Asymptotic statistical theory is a challenging domain for AI-assisted formalization: its central results mix convergence statements, asymptotic expansions, functional analysis, and regularity conditions that have a large gap from existing infrastructure in Lean 4 formalization. To address these challenges, we propose a hypothesis-disciplined Lean 4 formalization pipeline built from multiple agents: a manager that coordinates seven specialist roles for proof planning, skeleton scaffolding, Mathlib reconnaissance, proof construction, integration, independent review, and audit. The main methodological discipline is the hypothesis-disciplined audit, implemented by the Auditor agent: every main-theorem hypothesis and concept-layer field must be anchored in the source mathematical prose, justified as a Lean encoding adapter, marked as source-implied, or rejected as an unsupported strengthening. Using this workflow, we build a systematic formalization of asymptotic statistical theory, especially the parametric and semi-parametric models' asymptotic distribution and efficiency results. The resulting Lean development is axiom-clean and source-faithful, with Lean-checked and human-audited proofs of core parametric and semi-parametric theorems organized so that theorem-agnostic infrastructure and statistical concept definitions are separated from theorem-specific assembly. The formalization results are available at https://github.com/junwei-lu/Lean-Asymptotic-Statistical-Theory.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20640unread
An LLM-Explainable DRL Framework for Passenger-Directed Autonomous Driving
Ouided Braoui, Meriem Bouali, Nadir Farhi · 2026-06-23
arXiv:2606. 20640v1 Announce Type: new Abstract: Autonomous vehicles offer the potential for safer and more efficient mobility, yet public trust remains limited due to the lack of transparency in their decision-making.
Read next because An LLM-Explainable DRL Framework for Passenger-Directed Autonomous Driving 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, rate, control, trained, language, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20640v1 Announce Type: new Abstract: Autonomous vehicles offer the potential for safer and more efficient mobility, yet public trust remains limited due to the lack of transparency in their decision-making. This work addresses this issue by combining deep reinforcement learning (DRL) for adaptive driving control with large language model (LLM)-based explainability modules designed to communicate agent behavior to passengers. DRL agents were trained in simulation using a Dueling Double Deep Q-Network to follow distinct driving requests: \textit{fast}, \textit{comfort}, and \textit{stop}. They demonstrated stable learning, safe compliance with traffic rules, and reliable switching between modes within a single trip. In parallel, LLM modules were introduced to interpret passenger requests, determine when explanations were needed, and generate concise, safety-oriented justifications. Results show that this framework, serving as a proof of concept for integrating RL decision-making and LLMs, balances safety, adaptability, and explainability, and is most effective when requests are delayed or overridden due to safety constraints.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20631unread
Harnessing Agent Skills: Architectural Patterns and a Reference Architecture for Skill-Mediated LLM Agents
Boming Xia, Liming Zhu, Zhenchang Xing, Qinghua Lu, Dino Sejdinovic, Xiwei Xu · 2026-06-23
arXiv:2606. 20631v1 Announce Type: new Abstract: Agent skills externalise reusable agent-facing behavioural knowledge and guidance as persistent artefacts that can be discovered, activated, and interpreted by LLM agents.
Read next because Harnessing Agent Skills: Architectural Patterns and a Reference Architecture for Skill-Mediated LLM 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, eval, control, chain. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20631v1 Announce Type: new Abstract: Agent skills externalise reusable agent-facing behavioural knowledge and guidance as persistent artefacts that can be discovered, activated, and interpreted by LLM agents. Although a skill artefact is static at rest, its architectural responsibilities arise in use, when the artefact is selected for a run, bound to context and authority constraints, interpreted by a stochastic agent, and recorded as run evidence. We call this run-specific relation skill-in-use. This paper studies agent skill harnessing: the architectural responsibilities that govern the transition from skill artefacts to skill-in-use, bound the executable consequences associated with skill-in-use, and capture evidence for attribution, verification, repair, and evolution. This paper provides a catalogue of ten empirically grounded architectural patterns (five core, five supporting) for skill harnessing and synthesises them into a reference architecture with four responsibility layers: Supply Chain, Mediation, Execution Control, and Evidence & Feedback. We evaluate the architecture through cross-instantiation across 8 selected systems. The resulting patterns and reference architecture provide a vocabulary and diagnostic frame for analysing skill-harnessing responsibilities across agent systems.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20628unread
Human Decision-Making with AI Assistance under Correlated Features
Yanru Guan, Naveen Raman, Fei Fang · 2026-06-23
arXiv:2606. 20628v1 Announce Type: new Abstract: Humans increasingly make decisions with AI assistance; for example, doctors may follow AI-recommended diagnostic tests and base their diagnoses on the results.
Read next because Human Decision-Making with AI Assistance under Correlated Features 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, under, rate, length, suffix, test, lora. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20628v1 Announce Type: new Abstract: Humans increasingly make decisions with AI assistance; for example, doctors may follow AI-recommended diagnostic tests and base their diagnoses on the results. A natural question is which tests should AI recommend to balance short-term decision quality and long-term human learning when different features (e.g., test results) are correlated. While prior work establishes that stationary policies that recommend the same tests repeatedly are optimal when features are independent, we prove that feature correlations lead such policies to perform arbitrarily poorly. Instead, we prove that any optimal policy must follow an explore-then-commit structure; initially, the AI should offer diverse tests so humans can learn accurate feature coefficients, then the AI should commit to a single set of tests, with exploration length that depends on the degree of feature correlation. We prove that computing the optimal policy is NP-hard and derive a dynamic programming-based algorithm that finds the optimal policy for finite horizons. We additionally develop an approximation that plans for shorter horizons and appends a stationary suffix, achieving near-optimal performance. Our empirical results complement our theory by showing that stronger feature correlation leads to longer exploration phases.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20623unread
Path-dependent program induction under resource constraints explains human sequence learning
Hanqi Zhou, David G. Nagy, Peter Dayan, Charley M. Wu · 2026-06-23
arXiv:2606. 20623v1 Announce Type: new Abstract: How do people build abstract, reusable knowledge from sequential experience under bounded cognitive resources?
Read next because Path-dependent program induction under resource constraints explains human sequence 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: code, strong, under, source, line, rate, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20623v1 Announce Type: new Abstract: How do people build abstract, reusable knowledge from sequential experience under bounded cognitive resources? To answer this question, we integrate rate-distortion theory with recent advances in program induction to describe how prior knowledge shapes which future structures are cheap to encode and easy to discover. We formalize this in a hierarchical Adaptor Grammar (HAG) with distinct local (within-task) and global (across-task) libraries, governed jointly by constraints on memory and computation. In simulations, HAG achieves better rate-distortion trade-offs and stronger generalization than fixed grammars or shallow chunking methods. In an online melodic sequence-learning experiment, participants' recall errors reflected systematic simplifications and reaction times increased at inferred program boundaries. Trial-by-trial fits further showed that hierarchical libraries best explained individual differences in both recall and out-of-sample continuation choices, outperforming all alternative models. These findings cast structured learning as bounded program induction in which the order of experience shapes future abstractions a learner builds.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20600unread
The New Associationism: Lessons from Deep Learning
Daniel Rothschild · 2026-06-23
arXiv:2606. 20600v1 Announce Type: new Abstract: What can the success of modern AI tell us about how humans learn?
Read next because The New Associationism: Lessons from Deep Learning 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, eval, rate, language, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20600v1 Announce Type: new Abstract: What can the success of modern AI tell us about how humans learn? This paper argues that taking AI seriously as a model of human learning supports a modest but genuine associationism. The central finding is that supervised learning -- learning driven by evaluative feedback -- underlies a surprisingly wide range of contemporary AI systems, from large language models to game-playing agents, differing primarily in how much work is required to generate the relevant feedback signal. This vindicates associationist ideals of a uniform, gradual, error-driven learning mechanism operating across domains, and defuses the once-influential argument that associationist mechanisms are too limited to account for human cognitive capacities. At the same time, the successes of deep learning depend on computational architectures that go well beyond anything classical associationists envisaged, and supervised learning operates within these as one component rather than a complete account of learning.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20569unread
On the Identifiability of User Adaptation in Co-Adaptive Neural Interfaces
Philip Waggoner · 2026-06-23
arXiv:2606. 20569v1 Announce Type: new Abstract: We analyze identifiability in co-adaptive human-machine systems.
Read next because On the Identifiability of User Adaptation in Co-Adaptive Neural Interfaces 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: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20569v1 Announce Type: new Abstract: We analyze identifiability in co-adaptive human-machine systems. We show that closed-loop encoder estimates do not uniquely identify user adaptation, but instead reflect properties of the joint system. We discuss implications for interpreting behavioral adaptation and propose conditions for identification.
- score 100arxiv cs.CL (NLP)arxiv:2606.21075unread
FiLM-Coordinated Dual-Branch Transformer for Global-Local Dependency Modeling in Language Modeling
Zhiqiang Zhou, Xu Ling, Junliang Dai · 2026-06-23
arXiv:2606. 21075v1 Announce Type: new Abstract: Standard Transformers use a single self-attention pathway to model both global dependencies and local patterns, creating tension between long-range structural reasoning and fine-grained local representation learning.
Read next because FiLM-Coordinated Dual-Branch Transformer for Global-Local Dependency Modeling in Language 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: text, rect, under, width, token, line, rate, full. Source: arxiv cs.CL (NLP).
arXiv:2606.21075v1 Announce Type: new Abstract: Standard Transformers use a single self-attention pathway to model both global dependencies and local patterns, creating tension between long-range structural reasoning and fine-grained local representation learning. We propose a FiLM-coordinated dual-branch Transformer for language modeling, where each layer explicitly contains a global branch and a local branch, and feature-wise linear modulation (FiLM) is used for dynamic cross-branch coordination instead of simple concatenation or static addition. The key idea is that the two branches represent different dependency views of the same input, making channel-wise calibration more suitable than heavy token-level interaction. We therefore design a bidirectional FiLM module in which each branch generates per-channel scaling and shifting parameters to condition the other. Experiments on multiple small-scale language modeling settings show that the proposed structure consistently outperforms same-width single-branch baselines and weakened dual-branch variants under a fixed lightweight configuration. On TinyShakespeare and a 1M-character subset of WikiText-2, the full dual-branch FiLM model achieves the best results among same-width structural baselines. Multi-seed results support the stability of the gains, while mechanistic analyses show that FiLM learns input-dependent, layer-dependent, and channel-selective modulation patterns rather than static scaling. Parameter-matched widened single-branch baselines also indicate that the current design still leaves room for improvement in parameter efficiency.
- score 100arxiv cs.CL (NLP)arxiv:2606.21048unread
Event Ontology Expansion via LLM-Based Conceptualization
Weicheng Ren, Zixuan Li, Long Bai, Xiaolong Jin, Jiafeng Guo, Xueqi Cheng · 2026-06-23
arXiv:2606. 21048v1 Announce Type: new Abstract: Event ontology expansion aims to discover emerging event types from data and extend them to appropriate positions in the existing event ontology..
Read next because Event Ontology Expansion via LLM-Based Conceptualization 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, rate, position, language. Source: arxiv cs.CL (NLP).
arXiv:2606.21048v1 Announce Type: new Abstract: Event ontology expansion aims to discover emerging event types from data and extend them to appropriate positions in the existing event ontology.. Existing methods typically cluster contextualized trigger representations and attach induced clusters to the ontology based on instance-level similarity. However, ontology expansion requires concept-level semantics that characterize event types, whereas contextualized trigger representations often conflate these semantics with surface contextual variation, leading to unstable clustering and unreliable hierarchy expansion. To address this issue, we propose ConceptE, a conceptualization-enhanced framework for event ontology expansion. ConceptE first derives concept-level semantics by prompting an LLM with the sentence and event trigger, producing a concise concept name and a natural-language description. It then jointly encodes these semantics with trigger information to build concept-enhanced representations aligned with ontology-level reasoning. This representation design supports more coherent event clustering, more reliable hierarchy expansion, and ontology-consistent type naming. Experiments on ACE, ERE, and MAVEN demonstrate that ConceptE consistently outperforms state-of-the-art approaches across all subtasks of event ontology expansion. In particular, it achieves improvements of up to 12.37\% in BCubed-F1 for event clustering and 6.48\% in Taxo\_F1 for hierarchy expansion, demonstrating the effectiveness of the proposed ConceptE method.
- score 100arxiv cs.CL (NLP)arxiv:2606.20993unread
Phonemes to the Rescue: Multilingual Tokenization Based on International Phonetic Alphabet
Milan Mileti\'c, Julie Kallini, Ekaterina Shutova · 2026-06-23
arXiv:2606. 20993v1 Announce Type: new Abstract: Multilingual language models often exhibit performance disparities across languages that can arise as early as the tokenization stage.
Read next because Phonemes to the Rescue: Multilingual Tokenization Based on International Phonetic Alphabet 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, word, latin, alpha, source, token, rate, stage. Source: arxiv cs.CL (NLP).
arXiv:2606.20993v1 Announce Type: new Abstract: Multilingual language models often exhibit performance disparities across languages that can arise as early as the tokenization stage. Widely-used subword tokenization approaches favor high-resource languages, and tokenizer-free methods still yield longer sequences for scripts with a higher bytes-per-character ratio. To address these shortcomings, we propose to use the International Phonetic Alphabet (IPA) as a language-agnostic input representation for multilingual tokenizers. IPA provides a compact symbol inventory, greater cross-lingual character overlap, and a more balanced byte-per-character distribution across languages. We train matched pairs of text vs. IPA subword tokenizers across 24 languages and 14 scripts and demonstrate that IPA tokenizers consistently improve tokenization quality, especially for non-Latin scripts, and generalize more effectively to unseen languages and scripts.
- score 100arxiv cs.CL (NLP)arxiv:2606.20929unread
Peeking Inside LLMs: Leveraging Internal Artifacts of LLMs for Enhancing Reliability in Legal Classification
Sudipta Santra, Debtanu Datta, Saptarshi Ghosh · 2026-06-23
arXiv:2606. 20929v1 Announce Type: new Abstract: Large Language Models (LLMs) are increasingly being adopted in the legal domain.
Read next because Peeking Inside LLMs: Leveraging Internal Artifacts of LLMs for Enhancing Reliability in Legal 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: strong, class, rect, correct, eval, rate, language, model. Source: arxiv cs.CL (NLP).
arXiv:2606.20929v1 Announce Type: new Abstract: Large Language Models (LLMs) are increasingly being adopted in the legal domain. However, despite their strong performance, LLMs are prone to generating incorrect or hallucinated outputs, raising serious concerns about their reliability in high-stakes domains such as law. Detecting the correctness of responses of LLM-based systems is therefore a critical challenge. In this work, we explore the potential of leveraging internal artifacts of LLM to detect the correctness of their predictions in legal-domain classification tasks. We develop approaches that utilize features derived from these internal artifacts to build downstream classifiers capable of identifying incorrect LLM outputs. We evaluate our approach on two representative legal classification tasks: bail decision prediction and statute violation prediction. Our experimental results demonstrate that LLMs' internal artifacts are reliable indicators for detecting incorrect predictions in legal classification tasks, and can be applied to enhance the reliability of LLM-based classification systems.
- score 100arxiv cs.CL (NLP)arxiv:2606.20890unread
Topic-to-Timestamp Alignment by Constrained Evidence Selection
Zeynep Y{\i}lb{\i}rt, Marina Litvak, Michael F\"arber · 2026-06-23
arXiv:2606. 20890v1 Announce Type: new Abstract: Meeting archives are difficult to search when users remember what was discussed but not when.
Read next because Topic-to-Timestamp Alignment by Constrained Evidence 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, strong, alignment, eval, rate, trained, candidate, language. Source: arxiv cs.CL (NLP).
arXiv:2606.20890v1 Announce Type: new Abstract: Meeting archives are difficult to search when users remember what was discussed but not when. We study topic-to-timestamp alignment: given a natural-language topic and a timestamped meeting transcript, the goal is to return the time at which the topic is discussed. A standard RAG setup can retrieve relevant transcript excerpts, but still asks the language model to generate a timestamp, which can produce unsupported or invalid timecodes. We therefore recast timestamp prediction as constrained temporal candidate selection: the system retrieves timestamped transcript chunks, and the model selects the candidate that best grounds the topic instead of generating a timecode. On 420 topic-timestamp queries from 200 municipal meeting transcripts, this increases Recall@5 from 31.9% to 50.0%, reduces MAE from 837.0 seconds to 761.0 seconds with Mistral-7B-Instruct, and increases the number of parseable outputs from 373 to 419 of 420 queries. The results suggest that temporal grounding in long transcripts depends strongly on retrieval quality and output design, not only on the choice of the language model.
- score 100arxiv cs.CL (NLP)arxiv:2606.20873unread
SciLens: Multi-modal Scientific Claim Verification with Agentic Entailment and Grounding
Yueming Wang, Tianshi Zheng, Jiaxin Bai, Yangqiu Song, Ginny Wong, Simon See · 2026-06-23
arXiv:2606. 20873v1 Announce Type: new Abstract: Scientific discovery increasingly relies on automated systems that generate hypotheses, inspect multimodal evidence, and validate claims at scale.
Read next because SciLens: Multi-modal Scientific Claim Verification with Agentic Entailment and Grounding 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, eval, rate, language, model. Source: arxiv cs.CL (NLP).
arXiv:2606.20873v1 Announce Type: new Abstract: Scientific discovery increasingly relies on automated systems that generate hypotheses, inspect multimodal evidence, and validate claims at scale. Yet scientific claim verification is not well served by asking a vision-language model for a direct binary judgment: claims often combine numerical results, comparisons, scope qualifiers, and explanatory context, while evidence is encoded in tables and figures with distinct grounding structures. We present SciLens, an evidence-conditioned atomic entailment framework for multimodal scientific claim verification. SciLens decomposes each claim into central empirical atoms and background atoms, grounds the central atoms to modality-specific evidence witnesses, and predicts the final label with an atom-level entailment rule. For tables, atoms are grounded to rows, columns, cells, arithmetic relations, and table scope; for figures, they are grounded through panels, axes, legends, visual encodings, categories, trends, ranks, and qualifier checks. This yields a unified validation procedure in which a claim is supported only if every central empirical atom is entailed by the current evidence. On the SciClaimEval development set, SciLens achieves 79.2% macro-F1 and 63.1% pair accuracy, showing that structured agentic validation improves both evidence sensitivity and interpretability.
- score 100arxiv cs.CL (NLP)arxiv:2606.20769unread
FirstPass: Grounding AI Scientific Judgment in Multi-Round Editorial Outcomes
Prabhjot Singh, Somnath Luitel, Manmeet Singh, Josh Durkee · 2026-06-23
arXiv:2606. 20769v1 Announce Type: new Abstract: AI systems for peer review fail on three fronts: they train on Computer Science and Machine Learning venues alone, ignore the iterative dialogue that validates science, and evaluate on stylistic mimicry rather than real editorial judgment.
Read next because FirstPass: Grounding AI Scientific Judgment in Multi-Round Editorial Outcomes 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 "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: word, eval, line, without, alone, lora, qwen2, model. Source: arxiv cs.CL (NLP).
arXiv:2606.20769v1 Announce Type: new Abstract: AI systems for peer review fail on three fronts: they train on Computer Science and Machine Learning venues alone, ignore the iterative dialogue that validates science, and evaluate on stylistic mimicry rather than real editorial judgment. We introduce FirstPass, a dataset and fine-tuned model that addresses all three. Curating 3,668 complete multi-round peer-review dialogues from Nature Communications across five scientific domains (biology, chemistry, neuroscience, physics, and earth science), we exploit mandatory transparent peer review (instituted November 2022) and verify 100% content integrity by automated audit. We fine-tune Qwen2.5-7B-Instruct via Low-Rank Adaptation (LoRA) on three tasks: review generation, reviewer updating, and revision-cycle prediction. Our key finding is that response-only loss masking is a prerequisite, not an optimization: without it, accuracy is 62.0%, below the majority baseline; with it, FirstPass achieves 80.5% accuracy and F1-macro 78.2% on predicting editorial outcomes (Standard vs. Extended revision cycles), outperforming Gemini-3.1-flash-lite-preview zero-shot by 10.4 percentage points and all baselines with statistical significance (McNemar p < 0.001). On generation, FirstPass produces reviews averaging 1,187 words, substantially closer to human references (2,155 words) than any baseline, achieving ROUGE-L 0.154 with significant gains over Qwen and DeepSeek zero-shot (p < 0.001). Deployed in the pre-submission loop as an anticipatory scientific co-author, FirstPass simulates expert critique and predicts revision cycle outcomes before submission, giving authors the judgment a trusted colleague would provide, with consistent cross-domain performance across five disciplines.
- score 100arxiv cs.CL (NLP)arxiv:2606.20751unread
From Sentiment to Actionable Insights: A Data-Driven Public Sentiment Analysis of Advanced Air Mobility
Esrat Farhana Dulia, Amina Dhaher, Raiful Hasan, Syed Arbab Mohd Shihab · 2026-06-23
arXiv:2606. 20751v1 Announce Type: new Abstract: Advanced Air Mobility (AAM) is an emerging low-altitude air transportation system whose successful deployment depends not only on technological advancement but also on public acceptance.
Read next because From Sentiment to Actionable Insights: A Data-Driven Public Sentiment Analysis of Advanced Air Mobility 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, rate, full, model. Source: arxiv cs.CL (NLP).
arXiv:2606.20751v1 Announce Type: new Abstract: Advanced Air Mobility (AAM) is an emerging low-altitude air transportation system whose successful deployment depends not only on technological advancement but also on public acceptance. This acceptance will drive government support, regulations, noise standards, and willingness to fly, and in turn the overall commercial viability of AAM. Understanding public sentiment toward AAM is therefore essential for identifying its societal barriers and informing its adoption strategies. This study analyzes 306,009 human-generated texts collected from Reddit and Quora to examine public discourse on AAM using AI-based models. Because multiple sentiment analysis models exist, identifying the most accurate model is critical for reliable AAM sentiment prediction and trustworthy public opinion analysis. Accordingly, seven models spanning lexicon-based, machine learning, deep learning, and transformer-based approaches are evaluated for AAM-specific sentiment classification. ModernBERT achieves the best classification performance and is used to label the full dataset. Using the resulting sentiment labels, Latent Dirichlet Allocation (LDA) is applied within each sentiment class to uncover latent topics in public opinion. The analysis identifies 20 distinct topics and traces their temporal evolution from 2008 to 2025. A cross-sentiment topic analysis further reveals six major clusters of public concern: workforce and skill development (25.29% of the dataset), regulation and compliance (24.64%), technical performance of drones (20.99%), military, geopolitics, and defense (14.58%), safety and operational risks (8.51%), and noise and disturbance (5.98%). Based on these findings, this study provides actionable strategies to address these concerns, thereby, improving public acceptance and support AAM deployment.
- score 100arxiv cs.CL (NLP)arxiv:2606.20696unread
MindAlign: Decoding Inner Speech from fMRI Signals via Multimodal Embedding Alignment under Limited Data
Muxuan Liu, Ichiro Kobayashi, Satoshi Nishida · 2026-06-23
arXiv:2606. 20696v1 Announce Type: new Abstract: Decoding inner speech from non-invasive brain signals remains a fundamental challenge due to the absence of overt linguistic output, limited training data, and large inter-subject variability.
Read next because MindAlign: Decoding Inner Speech from fMRI Signals via Multimodal Embedding Alignment under Limited 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, rect, under, alignment, line, rate, project. Source: arxiv cs.CL (NLP).
arXiv:2606.20696v1 Announce Type: new Abstract: Decoding inner speech from non-invasive brain signals remains a fundamental challenge due to the absence of overt linguistic output, limited training data, and large inter-subject variability. Existing brain-to-text approaches often rely on task-specific decoder fine-tuning, which restricts scalability and complicates adaptation to new participants. We propose MindAlign, a decoupled two-stage brain-to-language framework that enables open-ended text generation from fMRI signals without modifying the underlying language model. The first stage learns a subject-specific neural-semantic alignment that maps fMRI activity into a shared multimodal semantic space, extracting a latent semantic sketch of the internally generated sentence. The second stage integrates this sketch with visual context to prompt a frozen multimodal language model for free-form generation. Experiments on fMRI data collected during silent image description demonstrate that the proposed approach consistently outperforms fMRI-only and random baselines. We further show that the learned semantic-to-language projection can generalize across subjects, enabling effective decoding when paired with subject-specific neural alignment. These results indicate that neural signals modulate semantic content beyond image-driven priors, supporting a scalable and modular direction for brain-to-text decoding.
- score 100arxiv cs.CL (NLP)arxiv:2606.20691unread
Specific Domain Ontology Construction Using Large Language Models
Vivian Magri Alcaldi Soares, Renata Wassermann · 2026-06-23
arXiv:2606. 20691v1 Announce Type: new Abstract: Ontologies are useful structures to organize and maintain information that can be understood both by humans and systems.
Read next because Specific Domain Ontology Construction Using 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: text, under, eval, rate, without, factor, language, model. Source: arxiv cs.CL (NLP).
arXiv:2606.20691v1 Announce Type: new Abstract: Ontologies are useful structures to organize and maintain information that can be understood both by humans and systems. However, since their manual crafting is a laborious task, many specific domains lack reference ontologies. The outstanding ability for understanding natural language demonstrated by the Large Language Models (LLMs) has motivated their application to aid on a variety of fields, including on ontology development. This work presents the experimentation with a technique that uses LLMs in the role of domain experts to build conceptual hierarchies for a given initial concept. Twenty ontologies automatically constructed for the domain of the Brazilian maritime territory (a.k.a the Blue Amazon) using GPT-3.5 and GPT-4 were then evaluated by human experts. The models were able to construct overall coherent conceptualizations of the domain, but none of the outputs was completely satisfactory as a representation of the context without refinement.
- score 100arxiv cs.CL (NLP)arxiv:2606.20650unread
EmoInstruct-TTS: Dual-Path Instruction-Guided Emotional Speech Synthesis
Minghui Wu, Ganjun Liu, Zikun Fang, Ting Meng, Hongchuan Wu, Bingao Xu, Yonglong Cai, Jiasheng Chen, Jun Du · 2026-06-23
arXiv:2606. 20650v1 Announce Type: new Abstract: Instruction-based controllable speech synthesis enables users to specify emotions through natural language.
Read next because EmoInstruct-TTS: Dual-Path Instruction-Guided Emotional Speech Synthesis 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, language, model. Source: arxiv cs.CL (NLP).
arXiv:2606.20650v1 Announce Type: new Abstract: Instruction-based controllable speech synthesis enables users to specify emotions through natural language. However, existing approaches often rely on coarse emotion labels and lack explicit modeling of fine-grained intensity. We propose EmoInstruct-TTS, a dual-path instruction-guided framework for emotional speech synthesis. We introduce Emotion2embed, a supervised semantic-acoustic emotion embedding covering 48 emotional states, including fine-grained categories and intensity levels. To infer embeddings from free-form instructions, we design an Instruction-Conditioned Emotion Flow Model (ICE-Flow) that generates acoustically grounded emotion representations. The inferred embeddings are integrated into an LLM-based synthesis pipeline to provide explicit emotional control while preserving semantic planning. Experiments show improved emotional controllability and speech naturalness over strong baselines.
- score 100arxiv cs.CL (NLP)arxiv:2606.20632unread
Post-Training Recipe, More Than Model Family, Shapes Multi-Agent LLM Conversational Behavior
Luyang Zhang, Jialu Wang, Fei Xue, Yi-Yun Chu · 2026-06-23
arXiv:2606. 20632v1 Announce Type: new Abstract: Multi-LLM systems use multiple language models to deliberate, judge each other's outputs, or coordinate as agents.
Read next because Post-Training Recipe, More Than Model Family, Shapes Multi-Agent LLM Conversational Behavior 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, recipe, control, alone, chain, factor. Source: arxiv cs.CL (NLP).
arXiv:2606.20632v1 Announce Type: new Abstract: Multi-LLM systems use multiple language models to deliberate, judge each other's outputs, or coordinate as agents. Their value depends on the models producing measurably different conversational behaviors when given the same input. Prior offline studies recommend drawing one model per family for behavioral diversity, because LLMs prefer outputs from their own family when rating one another in isolation. Whether the same family label predicts behavior in interactive multi-LLM systems, the setting that real deployed systems use, has not been tested. We study this with a 940,000-chain 11-checkpoint corpus and a 1.6M-chain same-base Llama factorial. On our validated headline metric, hedging, a reasoning-distilled Llama checkpoint shifts by 18% depending on which same-base partner it replies to, more than any cross-family hedging gap in the controlled subset. Qwen, closed-API, and runtime checks suggest the pattern is not isolated, while repair and challenge analyses remain exploratory because their surface-cue detectors are weaker. Overall, the results identify post-training recipe as a first-class axis for multi-LLM panel composition and show that model family alone is an incomplete proxy for conversational diversity.
- score 100arxiv cs.CL (NLP)arxiv:2606.20571unread
Less is More: Lightweight Prompt Compression for Question Answering Applications on Edge Devices
Zihuai Xu, Ruofei Hou, Yang Xu, Hongli Xu, Yunming Liao, Ying Zhu · 2026-06-23
arXiv:2606. 20571v1 Announce Type: new Abstract: In agent-driven question answering (QA) applications, retrieval-augmented generation (RAG) is commonly introduced to enhance the response accuracy of large language models (LLMs) by providing additional context.
Read next because Less is More: Lightweight Prompt Compression for Question Answering Applications on Edge Devices 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, source, token, line, rate, implement, compare. Source: arxiv cs.CL (NLP).
arXiv:2606.20571v1 Announce Type: new Abstract: In agent-driven question answering (QA) applications, retrieval-augmented generation (RAG) is commonly introduced to enhance the response accuracy of large language models (LLMs) by providing additional context. Due to the inherent noise in retrieval results and the coarse granularity of document-level retrieval, the retrieved context often contains substantial redundant information. In this setting, the agent prompt, consisting of the user query and the associated retrieved context, leads to unnecessary computational overhead during LLM inference. Existing prompt compression methods typically rely on auxiliary small language models (SLMs) to estimate context importance. However, such approaches introduce significant memory and computational overhead, which limits their deployment on resource-constrained edge devices. In this paper, we propose CORE, a two-stage sentence-level prompt compression method that eliminates the need for SLMs. In the first stage, CORE constructs an answer set via named entity recognition (NER) and a clue set via semantic matching. In the second stage, CORE refines the clue set using an orthogonal residual retrieval strategy and designs a spatial proximity-based metric to filter the answer set. The two sets are then combined to form the final compressed context. We implement CORE on an NVIDIA Jetson AGX Orin edge device and a Huawei Nova smartphone. Experimental results demonstrate that within a 2000-token budget, CORE improves accuracy by at least 30.19% compared to state-of-the-art baselines, while reducing memory usage by at least 50.47% and achieving at least 1.94 times speedup on the edge device. Moreover, compared to the state-of-the-art LLMLingua2 method, CORE achieves a substantial energy reduction of 95.74% on the smartphone, highlighting its practicality and generalizability for mobile deployments.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.21021unread
Continuous-Time Probabilistic Correctors for Uncertainty-Aware Physics-Based Spacecraft Trajectory Forecasting
Muhammad Bilal Shahid, Zhanhong Jiang, Soumik Sarkar, Cody Fleming · 2026-06-23
arXiv:2606. 21021v1 Announce Type: new Abstract: Long-horizon spacecraft trajectory forecasting suffers from error accumulation due to the absence of corrective observations in the forecast regime, making reliable uncertainty estimation crucial for safety-critical decision-making such as space domain awareness and conjunction assessment.
Read next because Continuous-Time Probabilistic Correctors for Uncertainty-Aware Physics-Based Spacecraft Trajectory Forecasting 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, line, rate, compare, control. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.21021v1 Announce Type: new Abstract: Long-horizon spacecraft trajectory forecasting suffers from error accumulation due to the absence of corrective observations in the forecast regime, making reliable uncertainty estimation crucial for safety-critical decision-making such as space domain awareness and conjunction assessment. While high-fidelity physics-based orbit propagators provide accurate deterministic forecasts, they typically lack calibrated uncertainty estimates over long horizons. We introduce a Predictor--Corrector framework in which a physics-based continuous-time $\textit{deterministic}$ forecaster is augmented with a learned continuous-time $\textit{probabilistic}$ Corrector that models forecast errors. The proposed Corrector can be wrapped around an existing deterministic propagator to improve forecast accuracy while producing sharp and calibrated full-covariance uncertainty estimates. The Corrector is based on Latent Neural Controlled Differential Equations (Latent NCDEs) and models the probabilistic temporal evolution of forecast errors in continuous time, naturally supporting irregular sampling and missing features. We further introduce a loss function that promotes calibration and sharpness in long-horizon uncertainty propagation. We evaluate the proposed framework on long-horizon spacecraft trajectory forecasting using real-world data from NASA's Crustal Dynamics Data Information System (CDDIS), wrapping the Corrector around NASA's General Mission Analysis Tool (GMAT). Across forecast horizons of 2--4 days without observations and six rolling test windows, the proposed approach consistently improves accuracy and uncertainty calibration compared to deterministic baselines and Latent ODE-based correctors, demonstrating the effectiveness of the continuous-time probabilistic Corrector for trajectory forecasting.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20967unread
Formalizing Task-Space Complexity for Zero-Shot Generalization
Jung-Hoon Cho, Heling Zhang, Siqi Du, Roy Dong, Cathy Wu · 2026-06-23
arXiv:2606. 20967v1 Announce Type: new Abstract: Policies must operate across diverse conditions, yet a single policy is often conservative while fully adaptive schemes can be complex.
Read next because Formalizing Task-Space Complexity for Zero-Shot Generalization 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, rect, under, source, line, rate, control. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20967v1 Announce Type: new Abstract: Policies must operate across diverse conditions, yet a single policy is often conservative while fully adaptive schemes can be complex. We study zero-shot generalization in contextual dynamical systems and introduce a performance-centric, directional task dissimilarity--the signed divergence--that upper bounds the generalization gap from a source context to a target context. The signed divergence induces $\varepsilon$-tolerance sets that certify when a source policy class generalizes, and it yields a concrete notion of task-space complexity: the minimum number of source contexts needed so that every target context incurs at most $\varepsilon$ generalization gap. Under a mild local smoothness assumption on performance, the induced tolerance sets admit certified inner/outer balls and instance-dependent volume bounds on task-space complexity. In the finite-oracle setting, source selection reduces to set cover; a greedy strategy inherits the standard $H(n)$ approximation guarantee. Using a Mass-Spring-Damper system with linear-quadratic regulator (LQR) controllers and a nonlinear CartPole system with deep reinforcement learning controllers, we show that greedy selection achieves the same $\varepsilon$-coverage with fewer policies than uniform or random baselines. Our approach delivers a performance-based task similarity measure and practical certificates for building generalizable control with simple policies.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20955unread
A Gated Graph Neural Network Approach to Fast-Convergent Dynamic Average Estimation
Antonio Marino, Claudio Pacchierotti, Paolo Robuffo Giordano · 2026-06-23
arXiv:2606. 20955v1 Announce Type: new Abstract: Dynamic average estimation is a critical problem in multi-agent systems, enabling agents to collaboratively estimate time-varying signals using only local information exchange.
Read next because A Gated Graph Neural Network Approach to Fast-Convergent Dynamic Average Estimation 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, compare, without, full, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20955v1 Announce Type: new Abstract: Dynamic average estimation is a critical problem in multi-agent systems, enabling agents to collaboratively estimate time-varying signals using only local information exchange. Traditional model-based approaches often face challenges related to convergence speed and sensitivity to network topology changes. This paper introduces a novel learning-based solution leveraging Gated Graph Neural Networks (GGNNs) for fast-convergent dynamic average estimation in a fully distributed manner. Taking advantage of the inherent structure of GGNNs, the proposed method models the estimation process as a distributed autoregressor, ensuring rapid convergence while maintaining stability. We incorporate a regularization term during training to enforce convergence guarantees and introduce an encoding-decoding mechanism to reduce communication overhead without sacrificing accuracy compared to standard GGNNs. Extensive numerical experiments demonstrate that our approach significantly outperforms conventional model-based estimators in terms of both convergence speed and precision, making it a promising alternative for multi-agent applications that require dynamic average estimation.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20945unread
Grouped Query Experts: Mixture-of-Experts on GQA Self-Attention
Vishesh Tripathi, Abhay Kumar · 2026-06-23
arXiv:2606. 20945v1 Announce Type: new Abstract: Self-attention is central to Transformer performance and is often the most expensive part of the Transformer at long context lengths because its pairwise token interactions scale quadratically with sequence length.
Read next because Grouped Query Experts: Mixture-of-Experts on GQA Self-Attention 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, token, line, length. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20945v1 Announce Type: new Abstract: Self-attention is central to Transformer performance and is often the most expensive part of the Transformer at long context lengths because its pairwise token interactions scale quadratically with sequence length. Standard dense attention also applies the same set of attention heads to every token regardless of token difficulty or information content. This uniform activation can waste compute, especially as sequences grow longer and attention cost increases rapidly. We propose Grouped Query Experts (GQE), a mixture-of-experts layer on top of grouped-query attention (GQA). Within each GQA group, a router selects k query-head experts per token while all key-value (KV) heads remain dense and unchanged. Thus, GQE keeps the KV cache benefits of GQA and reduces only the active query-head computation. On a fixed 30B token budget at the 250M parameter scale, GQE matches the all-active GQA baseline in downstream accuracy while activating half the query heads per token.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20937unread
Learning through Internalization
Nikolaos Tsilivis, Nirmit Joshi, Marko Medvedev, Julia Kempe, Nati Srebro · 2026-06-23
arXiv:2606. 20937v1 Announce Type: new Abstract: We study internalization processes, by which neural-network-based systems absorb an explicit computational procedure into their own weights, and how they facilitate learning.
Read next because Learning through Internalization 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, rect, token, without, chain. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20937v1 Announce Type: new Abstract: We study internalization processes, by which neural-network-based systems absorb an explicit computational procedure into their own weights, and how they facilitate learning. We investigate how transformers internalize the simulation of semiautomata by internalizing chain-of-thought (CoT) tokens, which classes of semiautomata are harder to internalize, and expose the flip side of internalization, that is, a progressive degradation of out-of-distribution performance. We then provide the first provable analysis of successful internalization: for the task of learning parities, we show that a simplified one-layer transformer provably first learns the target with explicit CoT supervision and then internalizes the autoregressive generation as CoT tokens are progressively removed, learning to directly compute the parity. This task is computationally hard to learn from data without CoT supervision. Finally, we discuss how learning through internalization relates to the \textit{Positive Distribution Shift} phenomenon recently introduced by~\citet{Med+26}.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20932unread
Hierarchical Pooling for Sheaf Neural Networks
Dionisia Naddeo, Carlo Abate, Pietro Li\`o, Nicola Toschi, Filippo Maria Bianchi · 2026-06-23
arXiv:2606. 20932v1 Announce Type: new Abstract: Sheaf Neural Networks (SNNs) generalize Graph Neural Networks (GNNs) by replacing scalar node signals with stalk-valued signals and by using restriction maps to measure compatibility across edges.
Read next because Hierarchical Pooling for Sheaf Neural Networks 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, implement, project, chain, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20932v1 Announce Type: new Abstract: Sheaf Neural Networks (SNNs) generalize Graph Neural Networks (GNNs) by replacing scalar node signals with stalk-valued signals and by using restriction maps to measure compatibility across edges. Unlike standard graph diffusion, which encourages neighboring node features to become similar, sheaf diffusion promotes consistency through the restriction maps and can therefore model more general relationships between neighboring nodes. However, existing sheaf neural architectures mainly operate at a fixed graph resolution and do not provide a principled pooling mechanism for building hierarchical representations. In this paper, we introduce Hierarchical Sheaf Pool (HiSP), a sheaf-aware pooling framework based on local spectral coarsening. Given a partition of the graph, HiSP constructs each coarse stalk by projecting fine stalk-valued features onto the low-frequency eigenmodes of the cluster-internal sheaf Laplacian. These local modes define a cochain-level prolongation map, which allows the fine sheaf energy to be represented on the coarse space through a Galerkin operator. We further analyze the approximation induced by coarsening by separating truncation loss, due to discarded local modes, from realization loss, due to representing the projected operator as a coarse sheaf. Finally, we implement HiSP as a GNN pooling layer compatible with SNNs and provide a PyG implementation supporting batching, lifted sheaf Laplacians, and hierarchical architectures.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20918unread
Short-Term Electricity Demand Forecasting for New England Using a Hybrid Transformer-XGBoost Framework with Weather, Calendar, and COVID-19 Indicators
Reza Ghanavati, Behrooz Mosallaei · 2026-06-23
arXiv:2606. 20918v1 Announce Type: new Abstract: Accurate short-term electricity demand forecasting is critical for reliable power system operation, energy market planning, and infrastructure optimization.
Read next because Short-Term Electricity Demand Forecasting for New England Using a Hybrid Transformer-XGBoost Framework with Weather, Calendar, and COVID-19 Indicators 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, correct, line, rate, extraction, leakage. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20918v1 Announce Type: new Abstract: Accurate short-term electricity demand forecasting is critical for reliable power system operation, energy market planning, and infrastructure optimization. This paper presents a hybrid framework combining a Transformer encoder for temporal feature extraction with gradient-boosted decision trees (XGBoost) for daily electricity demand forecasting across New England. The framework integrates meteorological observations from six cities spanning all six New England states, calendar and holiday effects, autoregressive demand lags, and COVID-19 epidemiological variables. Hyperparameter optimization uses Optuna with a multivariate Tree-structured Parzen Estimator over 500 trials, with a leakage-free 70/15/15 chronological train-validation-test split. The hybrid model achieves a test RMSE of 8,876 MWh, MAPE of 2.05%, and R-squared of 0.906. A tabular-only XGBoost baseline achieves RMSE of 9,304 MWh, MAPE of 2.21%, and R-squared of 0.896. A Diebold-Mariano test (Harvey-Leybourne-Newbold correction) confirms the 427.7 MWh difference is statistically indistinguishable from noise (DM = -1.126, p = 0.262). An ablation study reveals COVID-19 features improved training accuracy but had asymmetric test effects: removal degraded hybrid RMSE by 3.2% while marginally improving XGBoost-only by 1.2%. A SHAP temporal analysis shows 5 of 8 COVID features rank higher on the post-acute test set than during pandemic-active training, indicating the model over-applies learned pandemic patterns. These findings establish temporal validity decay as a central mechanism: behavioral disruptions drove a strong COVID-demand signal during 2020-2021, but adaptation was complete by mid-2022, leaving epidemiological features as noise amplifying overfitting to stale pandemic patterns.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20858unread
Evolutionary Discovery of Developmental Reward Schedules in Deep Reinforcement Learning
Alan Nadelsticher Ruvalcaba · 2026-06-23
arXiv:2606. 20858v1 Announce Type: new Abstract: The temporal structure of reward composition in reinforcement learning (RL) is typically hand-designed and held fixed throughout training, leaving the progression of motivational priorities largely unexplored.
Read next because Evolutionary Discovery of Developmental Reward Schedules in Deep Reinforcement 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: code, eval, line, compare, project, position, capability. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20858v1 Announce Type: new Abstract: The temporal structure of reward composition in reinforcement learning (RL) is typically hand-designed and held fixed throughout training, leaving the progression of motivational priorities largely unexplored. In this work, we propose an evolutionary framework for discovering developmental reward schedules, in which three distinct biologically inspired motivational components -- agency, novelty, and reactivity -- are combined through time-varying weights that dynamically shift over the course of training. Evaluated on two sparse-reward MiniGrid tasks: DoorKey-6x6 and KeyCorridorS3R1, our framework compares the generalizability of four evolutionary algorithms: CMA-ES, xNES, DE, and L-SHADE against an extrinsically motivated baseline (our main comparison point), and three additional hand-designed methods. On DoorKey-6x6, all evolved methods outperform the non-evolved baselines, with L-SHADE achieving the best performance -- an approximate relative mean improvement of 11.4% over the extrinsic only baseline. On KeyCorridorS3R1, CMA-ES achieves the best overall performance, with the remaining evolved methods showing weaker and less reliable generalization capability compared to the extrinsic only baseline. Interestingly, the discovered schedules diverge from our defined developmental ordering, with novelty consistently emerging as the dominant early signal during training, across both tasks. Collectively, our results position evolutionary optimization as a promising approach for developmental reward schedule discovery in deep reinforcement learning, and suggest that what evolution finds to be optimal in computational settings may differ from what it finds to be optimal in biology. The code for this project can be found at: https://github.com/alannadels/Evolutionary_RL.git.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20747unread
CIExplainer++: Generating Causal and Interpretable Explanations for Graph Neural Networks
Francisco Caldas, Sahil Satish Kumar, Ruben Belo, Cl\'audia Soares · 2026-06-23
arXiv:2606. 20747v1 Announce Type: new Abstract: Explainable Artificial Intelligence aims to make black-box models more trustworthy by presenting, in a human-understandable manner, the elements that lead to the model's output.
Read next because CIExplainer++: Generating Causal and Interpretable Explanations for Graph 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 "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: latin, under, eval, compare, language, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20747v1 Announce Type: new Abstract: Explainable Artificial Intelligence aims to make black-box models more trustworthy by presenting, in a human-understandable manner, the elements that lead to the model's output. This involves both (i) identifying components and connections with genuine causal influence on outputs and (ii) translating such structures into an interpretable representation. For the former, we introduce CIExplainer, a novel perturbation-based method grounded in causal inference for explaining Graph Neural Networks (GNNs). CIExplainer identifies the subgraph with the highest causal effects on GNN predictions using the Potential Outcome Framework. We evaluate and compare CIExplainer on various GNN architectures (GCN, GraphSAGE, GAT, GIN) and datasets. To bridge subgraph explanations with human interpretability, we further propose G2TeXplainer, a method that transforms causal subgraphs into natural language explanations that capture both feature-level and relational information.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20743unread
Massive Activations Are Architecturally Robust: A Controlled Scratch/Commitment Residual Stream Test
Maruthi Vemula (University of North Carolina at Chapel Hill) · 2026-06-23
arXiv:2606. 20743v1 Announce Type: new Abstract: Trained transformers reliably develop massive activations, a small number of hidden dimensions whose magnitude is far above the median and which concentrate on the sequence-start token.
Read next because Massive Activations Are Architecturally Robust: A Controlled Scratch/Commitment Residual Stream Test 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, token, rate, control, does, trained. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20743v1 Announce Type: new Abstract: Trained transformers reliably develop massive activations, a small number of hidden dimensions whose magnitude is far above the median and which concentrate on the sequence-start token. Whether these outliers are a removable artifact of the residual stream's overloaded read and write role, or instead a functional necessity, is actively debated. We test the artifact hypothesis directly, with an architectural intervention. Our architecture, Ledger Residuals, splits the residual stream into a mutable scratch stream (Deliberation) that intermediate computation may freely overwrite and a protected, decode-only accumulator (Commitment) that holds the representation the model reads out. If massive activations exist only because one stream is forced to be both scratchpad and answer, then a dedicated answer channel should remove the need for them. We find that it does not. In matched-loss language models at the 160M and 290M scales, the model rebuilds the canonical fixed-dimension, start-token outlier inside the protected channel. The rebuilt feature is smaller in magnitude than in a standard transformer but more sharply concentrated on the start token, and a stronger sparsity penalty makes it more persistent and more concentrated still, rather than removing it. Massive activations therefore look architecturally robust: they re-emerge in whichever representation the model decodes from, which is what we would expect if they are functional rather than incidental. We release our architecture and measurement code.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.20878unread
Understanding Latent Flow Models for Tabular Data Synthesis: Targets, Paths, and Sampling
Bahrul Ilmi Nasution · 2026-06-23
arXiv:2606. 20878v1 Announce Type: new Abstract: Synthetic tabular data enables microdata sharing in regulated domains, yet deploying continuous-time generative models requires balancing analytical utility, disclosure risk, and computational cost.
Read next because Understanding Latent Flow Models for Tabular Data Synthesis: Targets, Paths, and Sampling 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, distributional, eval, source, rate, implement, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20878v1 Announce Type: new Abstract: Synthetic tabular data enables microdata sharing in regulated domains, yet deploying continuous-time generative models requires balancing analytical utility, disclosure risk, and computational cost. Latent-space flow models are flexible, but theoretical equivalences across learning targets, probability paths, and sampling dynamics can translate into different behaviour under finite-step integration and explicit compute budgets. We present an empirical study of tabular latent flow models across seven datasets, evaluating velocity, score, noise, and posterior matching objectives under optimal transport (OT) and variance-preserving (VP) paths, ODE and SDE sampling, and varying integration budgets. Our contributions are threefold: (1) we show that the learning target largely determines the utility-risk operating regime, with velocity and posterior matching tending to yield higher utility, while score and noise matching tend to achieve lower disclosure risk; (2) we demonstrate that configuration and sampling choices shift performance, with midpoint often improving distributional fidelity and OT paths often tolerating earlier stopping than VP, enabling compute savings under fixed budgets or risk thresholds; and (3) we distil these findings into actionable defaults and practical configuration guidance to support pre-release model selection under disclosure risk and resource constraints. The code implementation and supplementary materials can be accessed in https://github.com/rulnasution/tabular-latent-flow/.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.19367unread
Weibull Weight-Scale Parameter Evolution under AdamW Training Dynamics
Tiexin Ding · 2026-06-23
arXiv:2606. 19367v1 Announce Type: cross Abstract: Building on a two-parameter Weibull framework for diagnosing transformer weight distributions, we study why the Weibull weight-scale parameter $\lambda$ grows, overshoots, and then relaxes during AdamW training.
Read next because Weibull Weight-Scale Parameter Evolution under AdamW Training Dynamics 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, alignment, line, control, follow-up, trained. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.19367v1 Announce Type: cross Abstract: Building on a two-parameter Weibull framework for diagnosing transformer weight distributions, we study why the Weibull weight-scale parameter $\lambda$ grows, overshoots, and then relaxes during AdamW training. We derive a leading-order three-force decomposition of the squared weight norm from the AdamW update: an alignment force measuring the correlation between weights and the adaptive update direction, an injection force from adaptive step magnitude, and a decay force from decoupled weight decay. On self-trained Pythia-70M models with ground-truth optimizer moments, alignment dominates the rise phase, contributing 88-94% of the absolute force budget across four random seeds and remaining robust to super-weight removal. Near saturation, alignment and decay approach balance, explaining the transition from weight-scale growth to relaxation. These force dynamics directly govern the squared-norm component underlying $\lambda(t)$; the remaining RMS-to-Weibull reconstruction offset is measurable and decomposes into bridge and integration components, totaling approximately 5-6% in densely sampled regions. To extend the analysis to real models where optimizer moments are unavailable, we introduce a spline displacement method that recovers the alignment force from sparse checkpoints with approximately 92-94% accuracy, about twice the naive two-point baseline. We further observe that the peak value of $\lambda(t)$ varies with training-data coherence in our experiments, suggesting a data-dependent component of weight-scale growth that we leave to a controlled follow-up study. Code and data are available at https://github.com/tiexinding/NPM-Weibull-public.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.23627unread
Diffusion Models Adapt to Low-Dimensional Structure Under Flexible Coefficient Choices
Changxiao Cai, Yuchen Jiao, Gen Li · 2026-06-23
arXiv:2606. 23627v1 Announce Type: new Abstract: Diffusion models are known to exploit unknown low-dimensional structure to accelerate sampling.
Read next because Diffusion Models Adapt to Low-Dimensional Structure Under Flexible Coefficient Choices 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, rate, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.23627v1 Announce Type: new Abstract: Diffusion models are known to exploit unknown low-dimensional structure to accelerate sampling. However, existing convergence theory under low-dimensional data structure has largely focused on update rules with narrowly prescribed coefficient choices. This raises a fundamental question: is adaptation to low-dimensional structure sensitive to the precise choice of update coefficients? In this paper, we show that such adaptation is a robust property of diffusion models. For a broad class of update coefficients, we prove that $\widetilde{O}(k/\varepsilon)$ iterations suffice to generate an $\varepsilon$-accurate sample in total variation (TV) distance, independently of the ambient dimension. Our framework substantially broadens the class of diffusion samplers known to enjoy low dimensional adaptation and applies to several commonly used methods in practice. These results provide a theoretical justification for the empirical effectiveness of diffusion samplers across different coefficient choices when applied to structured, high-dimensional data.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.23601unread
Neural Networks as Linear Regression: An Introduction for Statisticians
Abigail Loe, Susan Murray, Zhenke Wu · 2026-06-23
arXiv:2606. 23601v1 Announce Type: new Abstract: Neural networks are a commonly used prediction tool in computer science and statistics.
Read next because Neural Networks as Linear Regression: An Introduction for Statisticians 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, trained. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.23601v1 Announce Type: new Abstract: Neural networks are a commonly used prediction tool in computer science and statistics. However, the barrier to entry of this interesting field remains high, particularly for classical statisticians trained in a frequentist perspective. In this letter, we demystify neural networks by describing networks that approximate a linear regression and describe common customizations that provide a foundation for further study.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.23472unread
Time Series Classification through Diffeomorphic Time Warping (DiffTW)
Vicky Geneva Haney, Kamel Lahouel, Victor Rielly, Bruno M. Jedynak · 2026-06-23
arXiv:2606. 23472v1 Announce Type: new Abstract: Time series classification involves learning a mapping from a continuous, temporally ordered sequence of real-valued observations to a discrete response variable, like class labels.
Read next because Time Series Classification through Diffeomorphic Time Warping (DiffTW) 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, line, rate, control, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.23472v1 Announce Type: new Abstract: Time series classification involves learning a mapping from a continuous, temporally ordered sequence of real-valued observations to a discrete response variable, like class labels. This task is fundamental in domains, including health monitoring, where the temporal structure of data is critical for accurate prediction. Dynamic Time Warping (DTW) is a standard technique for measuring similarity between sequences varying in time or speed. However, DTW is restricted to discrete point matching. To move beyond pairwise alignment, we propose a theoretical framework that learns mappings between real-valued functions. These mappings approximate the flow associated with the characteristic curves of a linear transport equation with a space-dependent velocity field, providing a diffeomorphic transformation between two time series. Using the method of characteristics, we transform this partial differential equation into ordinary differential equations (ODEs) modeling system dynamics. The objective function used to learn these ODEs derives from the fundamental theorem of calculus. To enable flexible, expressive representations of the velocity field, we utilize reproducing kernel Hilbert spaces and optimal control methods. Our method, Diffeomorphic Time Warping (DiffTW), provides a theoretically grounded dissimilarity measure. Using a 1-nearest neighbor classifier, DiffTW outperforms DTW on 60 of 86 datasets.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.23047unread
Domain Adaptation Under Wireless Network Constraints: When Does It Become Green?
Illyyne Saffar, Aur\'elie Boisbunon, Shruti Bothe · 2026-06-23
arXiv:2606. 23047v1 Announce Type: new Abstract: The deployment of data-driven models in 6G wireless networks is increasingly challenged by frequent distribution shifts that degrade performance over time.
Read next because Domain Adaptation Under Wireless Network Constraints: When Does It Become Green? 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, compare, without, does, trained, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.23047v1 Announce Type: new Abstract: The deployment of data-driven models in 6G wireless networks is increasingly challenged by frequent distribution shifts that degrade performance over time. Unsupervised Domain Adaptation (UDA) offers an alternative approach by adapting the trained model to a shifted domain without requiring labels. However, UDA pipelines are often more complex than single-task training due to additional modules and optimization procedures, raising a practical question: do the benefits of adaptation come at a higher energy cost, and how does this trade-off compare to retraining when labeling effort is also considered? In this work, we investigate the energy consumption of UDA and compare it to single task. We further propose a way to determine the minimum number of target domains for which UDA becomes more energy-efficient than retraining, taking into account the labeling cost. Our results aim to clarify when UDA should be preferred over classical train-from-scratch approaches from an energy and labeling-aware perspective.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.22674unread
Data Evolution by Wittgenstein's Rule Following
Aydin Ghojogh, Benyamin Ghojogh · 2026-06-23
arXiv:2606. 22674v1 Announce Type: new Abstract: This paper introduces Wittgenstein's Rule Following (WRF) data evolution, a framework in philomatics for evolving or generating a new dataset from a sequence of previously observed datasets.
Read next because Data Evolution by Wittgenstein's Rule Following 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: latin, rect, distributional, rate, does, candidate. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.22674v1 Announce Type: new Abstract: This paper introduces Wittgenstein's Rule Following (WRF) data evolution, a framework in philomatics for evolving or generating a new dataset from a sequence of previously observed datasets. The method is inspired by Ludwig Wittgenstein's rule-following considerations and his notion of family resemblance in Philosophical Investigations. Unlike standard synthetic data generation, where the goal is usually to sample from or augment a fixed distribution, WRF aims to continue the implicit rule expressed by a historical sequence of datasets while preserving resemblance to the previous datasets. WRF represents each dataset by structural descriptors rather than pointwise correspondences. These descriptors summarize geometric, distributional, clustering, and, in the supervised case, label-based properties of the data. The method predicts a rule-following target by extrapolating descriptor trajectories and a family-resemblance target by averaging historical descriptors. Candidate datasets are then generated from the observed history through balanced or bounded mixture recombination, scored according to these targets, and optionally refined through differentiable optimization in descriptor space. The proposed framework allows both sample size and feature dimension to vary over time and does not assume that the next dataset is a direct transformation of the last one. Simulations on synthetic and image datasets show that WRF can generate meaningful continuations of evolving datasets in both unsupervised and supervised settings.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.22601unread
Scalable Bayesian Additive Models for Stellar Flare Detection via Amortized Gaussian Process Inference and Hidden Markov Models
Rodrigo Herrera, Vianey Leos-Barajas, Gwendolyn Eadie, Elizaveta Semenova, James Davenport · 2026-06-23
arXiv:2606. 22601v1 Announce Type: new Abstract: Gaussian Processes (GPs) are a powerful tool for Bayesian time-series modeling, yet their cubic computational cost remains a severe barrier for application to long, high-cadence datasets in astronomy.
Read next because Scalable Bayesian Additive Models for Stellar Flare Detection via Amortized Gaussian Process Inference and Hidden Markov 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, eval, line, rate, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.22601v1 Announce Type: new Abstract: Gaussian Processes (GPs) are a powerful tool for Bayesian time-series modeling, yet their cubic computational cost remains a severe barrier for application to long, high-cadence datasets in astronomy. While specialized scalable solvers like Celerite elegantly reduce this scaling to linear time, repeatedly evaluating the exact likelihood during iterative Bayesian sampling is a bottleneck for developing more complex models, like hierarchical or additive models in which Celerite is only one component. To make this inference computationally tractable, we introduce a generative surrogate framework. By utilizing a Variational Autoencoder (VAE) to learn a compressed representation of the Celerite prior, we map highly correlated stochastic dependencies into a low-dimensional, isotropic manifold. This transition completely bypasses exact covariance operations, shifting the computational burden to a rapid neural network forward pass. Through an extensive simulation study, we show that the generative surrogate accurately reproduces the structural fidelity of exact physical kernels like Celerite. Finally, we demonstrate embedding our VAE approximation into an additive model that combines Celerite and a hidden Markov model (HMM) for stellar flare detection in time series data of stars. We evaluate the joint VAE+HMM architecture against the exact Celerite+HMM framework on empirical astrophysical time series and demonstrate that the proposed methodology achieves significant reductions in computational time, enabling the rigorous, large-scale characterization of stellar flares across massive data archives.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.22336unread
Null-Calibrated Conformal Selection via Target-Membership Scores
Seungjin Choi · 2026-06-23
arXiv:2606. 22336v1 Announce Type: new Abstract: Conformal selection aims to identify test candidates whose unknown responses fall in a target region while controlling the false discovery rate.
Read next because Null-Calibrated Conformal Selection via Target-Membership Scores 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, under, rate, control, candidates, candidate, test. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.22336v1 Announce Type: new Abstract: Conformal selection aims to identify test candidates whose unknown responses fall in a target region while controlling the false discovery rate. Existing methods often inherit prediction-oriented nonconformity scores, such as residual or clipped residual scores, from conformal prediction. We argue that the natural score for selection is instead the target-membership probability. This score directly addresses the binary event being selected, and any monotone transform of it gives the Neyman--Pearson oracle ranking at a fixed null selection level. This distinction is irrelevant for mean-monotone targets, where conventional scores induce essentially the same ranking, but becomes important for interval-valued, variance-driven, multimodal, or multi-condition targets, where prediction-oriented scores can be misaligned with selection power. We study membership-score-based conformal selection and isolate one conformal calibration route, Null-Calibrated Conformal Selection (NCCS), which ranks test scores against confirmed non-target calibration examples. Under null exchangeability, NCCS yields finite-sample valid null p-values, which can be combined with BY under arbitrary dependence or with BH under standard positive-dependence conditions. Experiments support the score principle: membership scores match conventional scores on mean-monotone targets, substantially improve over mean-score selection on variance-driven targets, and, when calibrated by NCCS, trade power for finite-sample null validity in rare-target regimes where direct empirical-FDP thresholding can be anti-conservative.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.22259unread
Convergence Analysis of Nystr\"om Subsampling in Covariate Shift Adaptation for Misspecified case
Hanna Myleiko, Sergei Solodky, Vasyl Semenov · 2026-06-23
arXiv:2606. 22259v1 Announce Type: new Abstract: This paper investigates convergence properties of regularized Nystr\"om subsampling applied to the unsupervised domain adaptation problem under covariate shift.
Read next because Convergence Analysis of Nystr\"om Subsampling in Covariate Shift Adaptation for Misspecified case 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, rate, project. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.22259v1 Announce Type: new Abstract: This paper investigates convergence properties of regularized Nystr\"om subsampling applied to the unsupervised domain adaptation problem under covariate shift. We focus on the low-smoothness (misspecified) case where the target function lies outside the reproducing kernel Hilbert space. By combining Tikhonov regularization with Nystr\"om projection onto a subsampled subspace, we obtain upper bounds on the excess risk that hold with high probability and are expressed in terms of the source condition, the effective dimension, and the sample sizes. We further extend the analysis to the setting where the Radon-Nikodym derivative between the target and source marginal distributions is unknown and must be approximated, and we identify the minimal additional sample sizes required to maintain the same convergence rate as in the oracle case.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.22239unread
Variance-Tilted Diffusion Models for Diverse Sampling
Iskander Azangulov, Leo Zhang, Kianoosh Ashouritaklimi · 2026-06-23
arXiv:2606. 22239v1 Announce Type: new Abstract: Diffusion models are typically sampled independently, even when the downstream objective is to obtain a diverse set of candidates.
Read next because Variance-Tilted Diffusion Models for Diverse Sampling 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, candidates, candidate, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.22239v1 Announce Type: new Abstract: Diffusion models are typically sampled independently, even when the downstream objective is to obtain a diverse set of candidates. We introduce a variance-weighted batch distribution that favours collections of samples with large empirical spread after a prescribed linear feature map. The target is specified explicitly, and the sampler is derived as the corresponding Doob $h$-transform of independent diffusion dynamics. The resulting correction has a compact form: an interaction term that repels posterior denoised means, together with a curvature term that moves particles to the region of higher feature variance. This yields an interacting-particle sampler with a transparent probabilistic target rather than a heuristic repulsive drift.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.21875unread
Signed Evidence Flow: Conflict-Aware and Stability-Calibrated Data Analysis
Jeffery Opoku, David Banahene · 2026-06-23
arXiv:2606. 21875v1 Announce Type: new Abstract: Modern data analysis usually gives a prediction without showing whether the evidence behind it is clear, conflicting, or stable.
Read next because Signed Evidence Flow: Conflict-Aware and Stability-Calibrated Data Analysis 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, rate, without, position, test. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.21875v1 Announce Type: new Abstract: Modern data analysis usually gives a prediction without showing whether the evidence behind it is clear, conflicting, or stable. Two cases can have the same fitted confidence even when one has mostly agreeing evidence and the other has strong support and strong opposition. We propose Signed Evidence Flow (SEF), which combines a fitted prediction rule with signed feature attributions to measure support, opposition, conflict, and perturbation stability. We prove that confidence determines conflict exactly when it also determines total evidence mass, derive the remaining conditional variance, and state when conflict can improve loss prediction beyond confidence and other audit variables. We also connect conflict to geometric decision fragility. Across healthcare, Covertype, black-box, finance, and ten external data sets, conflict sometimes separates risk among predictions that already appear confident. Cross-fitted tests show added error-ranking information beyond confidence and attribution entropy on several data sets, including two large finance tasks. The direction is not universal: in some tasks, lowconflict cases are riskier. We therefore introduce ScopeGate, a held-out permutation diagnostic that checks the direction before SEF is used for review triage. SEF is consequently an audit tool rather than a universal risk score: it describes evidence structure, while an independent calibration sample determines whether that structure is useful in the target population.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.21080unread
Bayesian Model Averaging under Predictor Redundancy via Density-Ratio Posterior Compression
Hanqing Li, Xuewen Lu, Yuting Chen · 2026-06-23
arXiv:2606. 21080v1 Announce Type: new Abstract: Bayesian model averaging in support-indexed regression induces a posterior distribution over active predictor supports.
Read next because Bayesian Model Averaging under Predictor Redundancy via Density-Ratio Posterior Compression 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)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe". Matching terms: under, soft, compare, without, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.21080v1 Announce Type: new Abstract: Bayesian model averaging in support-indexed regression induces a posterior distribution over active predictor supports. Under predictor redundancy, posterior mass can spread across many nearly interchangeable supports, making exact-support summaries unstable or hard to interpret even when prediction is stable. We study how to report an already fitted Bayesian model averaging posterior without changing the Bayesian target. A report uses hard or soft regions of support space, and its compressed reporting law is compared with the reference posterior through an explicit density ratio. This ratio gives computable total-variation and Kullback--Leibler distortion, bounds for bounded predictive summaries, retained-mass diagnostics, and fallback-weight diagnostics. The framework covers fixed hard regions, metric-ball regions, posterior-cluster regions, and pooled-pruned region dictionaries. We prove exact error formulas and validation bounds for these region reports, and give conditions under which a few regions can replace a long list of individual supports. In simulations, our region reports often give shorter and clearer summaries while preserving the main posterior information, and the density-ratio diagnostics show when too much information has been lost.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.21036unread
Diffusion-Driven State Space Models
Jack Ruder, Michael Wojnowicz · 2026-06-23
arXiv:2606. 21036v1 Announce Type: new Abstract: In many domains, practitioners seek models that produce accurate forecasts while faithfully capturing latent system dynamics.
Read next because Diffusion-Driven State Space 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, under, rate, full, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.21036v1 Announce Type: new Abstract: In many domains, practitioners seek models that produce accurate forecasts while faithfully capturing latent system dynamics. Existing approaches typically sacrifice one of these goals: deep state space models often assume Gaussian latent transitions, limiting fit and forecasting, while diffusion models are highly expressive but lack principled inference for the underlying dynamics. To combine the strengths of both, we introduce the Diffusion-Driven State Space Model (DDSSM), which replaces the conventional Gaussian transition distribution with a diffusion model. Our DDSSM resolves the open problem of how to jointly train an autoencoder and a diffusion model on sequential data, thereby extending the literature on latent diffusion models for time series. Moreover, we find that the DDSSM empirically outperforms a state-of-the-art deep SSM at fitting and forecasting a simulated time series with multimodal transitions.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.20859unread
Betting on Moments: Legendre Jumper Martingales for Online Exchangeability Testing
Johan Hallberg Szabadv\'ary · 2026-06-23
arXiv:2606. 20859v1 Announce Type: new Abstract: We present a family of conformal test martingales based on shifted Legendre polynomials, which extends the Simple Jumper martingale.
Read next because Betting on Moments: Legendre Jumper Martingales for Online Exchangeability Testing 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, distributional, line, rate, factor, test. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.20859v1 Announce Type: new Abstract: We present a family of conformal test martingales based on shifted Legendre polynomials, which extends the Simple Jumper martingale. The Simple Legendre Jumper substitutes the linear betting function with a polynomial of arbitrary degree, thereby facilitating the detection of variance, skewness, and higher-order deviations from uniformity; the standard Simple Jumper is a specific instance of degree one. The Product Legendre Jumper integrates multiple polynomial degrees into a unified betting function, although its state space expands exponentially-a cost we refer to as the jumping tax. To address this issue, we introduce the Variational Legendre Jumper, which factorises the joint adaptation through a mean-field approximation, thereby reducing exponential scaling to linear time with minimal loss in power. Lastly, the Composite Legendre Jumper incorporates several jumping rates, ensuring a wealth floor under exchangeability and automatic adaptation to the shift's timescale. Empirical results from a real-world classification task demonstrate that the combined methods consistently surpass any single-degree martingale under distributional shift, and the composite variant is recommended as the default when the shift timescale is unknown.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.20678unread
Beyond Importance: Interchange-Sobol Sensitivity Reveals Task-Specific Content Channels in Transformer Components
Yifeng Guo, Jin-Hong Du, Xiang Chen · 2026-06-23
arXiv:2606. 20678v1 Announce Type: new Abstract: Mechanistic interpretability methods summarize a transformer component by a single importance score, conflating two distinct roles: a component may matter because it transports task-relevant content, or because the forward computation degrades when its contribution is removed.
Read next because Beyond Importance: Interchange-Sobol Sensitivity Reveals Task-Specific Content Channels in Transformer Components 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: donor, latin, under, eval, rate, compare, control, position. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.20678v1 Announce Type: new Abstract: Mechanistic interpretability methods summarize a transformer component by a single importance score, conflating two distinct roles: a component may matter because it transports task-relevant content, or because the forward computation degrades when its contribution is removed. We introduce \emph{Interchange-Group Sobol Decomposition} (IGSD), a paired-intervention framework that compares matched activation replacement with zero ablation on the same component, estimates two Sobol-style variance indices, and uses their signed difference to separate the two roles, with intervention validity monitored by a symmetric off-manifold diagnostic $\widehat{\mathrm{ST}}>1$. In factual recall, IGSD identifies an early-layer content channel in both GPT-2 small and Qwen2.5-1.5B that standard importance methods underestimate. A controlled subject and relation donor design shows that the early channel transports relation-frame content while late attention transports subject-retrieval content, refining at head granularity to the known $\mathrm{Attn}_{L9H8}$ head. Late-layer clamping confirms that the early signal is expressed through downstream transformations rather than residual pass-through. These results show that replacement and deletion are not interchangeable controls and their divergence provides a practical statistical diagnostic for content transport in transformer components.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21638unread
Toward Open Weight Models Without Risks: Separating Public and Private Capabilities in LLMs
Charbel El Feghali, Arkil Patel, Nicholas Meade, Spandana Gella, Verna Dankers, Siva Reddy · 2026-06-23
arXiv:2606. 21638v1 Announce Type: new Abstract: Open-weight Large Language Models (LLMs) enable scientific progress and broad deployment.
Read next because Toward Open Weight Models Without Risks: Separating Public and Private Capabilities in LLMs 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, extraction, control, without, capability, language, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21638v1 Announce Type: new Abstract: Open-weight Large Language Models (LLMs) enable scientific progress and broad deployment. However, they make it difficult to control access to sensitive capabilities. Current practice either suppresses dangerous capabilities before release or mediates access through closed services that use specialized model variants, input/output monitors, and API permissions. The former is susceptible to jailbreaks while sacrificing capability for all users to mitigate the risks posed by a few, and the latter is fundamentally incompatible with open-weight release. In this paper, we propose Tiered Language Models (TLMs), where a single set of released weights supports multiple capability levels. In its default public configuration, a TLM behaves as a conventional LLM. A compact secret key specifies a permutation over a small parameter subset, inducing an alternative computation graph over the same weights that exposes additional capabilities. We develop a training protocol that jointly pretrains both configurations from scratch, then fine-tunes the keyed configuration on private data with regularization to preserve the public model's behavior. We pretrain 180M- and 650M-parameter TLMs and demonstrate that the keyed configuration can acquire a new language, gain instruction-following ability, and memorize private factual knowledge, whereas the public configuration exhibits none of these capabilities. Moreover, we show that our approach extends naturally to multiple hierarchical tiers. Because authorization operates on the model's weight structure rather than in the input space, the mechanism resists fine-tuning-based extraction and partial key compromise. In general, TLMs take a step toward reconciling open-weight release with selective capability control.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21487unread
A Longitudinal Study of Android Apps Signing Key Protection
Mark Huasong Meng, Qing Zhang, Weirao Lu, Chunyang Chen · 2026-06-23
arXiv:2606. 21487v1 Announce Type: new Abstract: Android app signing relies on developer-managed credentials, making secure key protection essential for the integrity of the software supply chain.
Read next because A Longitudinal Study of Android Apps Signing Key Protection 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: word, under, soft, rate, chain, leakage, leaks, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21487v1 Announce Type: new Abstract: Android app signing relies on developer-managed credentials, making secure key protection essential for the integrity of the software supply chain. A recent platform key leakage incident involving two major OEM manufacturers demonstrates that even robustly designed signing mechanisms can be compromised due to developers' oversight. In this work, we conduct a longitudinal ecosystem study to characterize this threat by mining public repositories for Android signing credentials, recovering compromised keys via exposed passwords, and matching them against signatures from over 4,000 apps collected from major stores and OEM system images. Our analysis identifies 5,673 compromised keystores on GitHub and 26 unique certificates linked to 278 real-world apps. These include 26 third-party apps in public app stores and 252 preinstalled apps from seven manufacturers, collectively affecting over 10 billion users. We demonstrate the practical exploitability of these leaks through a proof-of-concept app replacement attack and identify spillover risks in non-smartphone platforms, including a popular automotive head-unit platform installed in over 1,100 vehicle models. Our results reveal that signing-key mismanagement is a systemic risk, underscoring the need for a more rigorous key-management support in Android release engineering and distribution infrastructures.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21349unread
LLM-assisted Generation of Pseudo-C2 Servers for IoT Malware Dynamic Analysis
K. Hasui, S. Matsugaya, M. Shimamura, M. Hashimoto · 2026-06-23
arXiv:2606. 21349v1 Announce Type: new Abstract: Most IoT malware operates as botnets dependent on Command and Control (C2) servers, but the short-lived nature of attack infrastructure often leaves samples dormant without C2 communication, hindering dynamic analysis.
Read next because LLM-assisted Generation of Pseudo-C2 Servers for IoT Malware Dynamic Analysis 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, rate, extraction, control, without, full, trained. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21349v1 Announce Type: new Abstract: Most IoT malware operates as botnets dependent on Command and Control (C2) servers, but the short-lived nature of attack infrastructure often leaves samples dormant without C2 communication, hindering dynamic analysis. This paper proposes a system that combines Ghidra with a Large Language Model (LLM) to extract communication specifications from a malware binary and automatically generate a pseudo-C2 server. Experiments using Mirai demonstrate that the proposed system semantically interprets binary control structures and extracts all 20 core protocol elements in agreement with the ground truth (100\% specification extraction accuracy). The generated pseudo-C2 server fully reproduces seven of ten DDoS attack vectors with attack behavior consistent with the original C2. When applied to a customized variant created by modifying the publicly available Mirai source code, the method succeeds end-to-end -- from specification extraction through pseudo-C2 generation to attack reproduction -- demonstrating that the LLM infers specifications from binary structures without relying on pre-trained knowledge. This approach extends the applicability of LLMs from analysis assistance to the automated construction of dynamic analysis environments.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21338unread
"What Happens Locally, Leaks Globally": Detecting Privacy Leakage Risks in MCP Servers
Biwei Yan, Minghui Xu, Yijun Yang, Boyang Ma, Xuelong Dai, Jingku Li, Yue Zhang · 2026-06-23
arXiv:2606. 21338v1 Announce Type: new Abstract: The Model Context Protocol (MCP) has rapidly become the de facto standard for connecting large language models (LLMs) to external resources, but it also introduces a class of privacy risks that existing tools are ill-equipped to detect.
Read next because "What Happens Locally, Leaks Globally": Detecting Privacy Leakage Risks in MCP 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, persona, class, source, token, rate, implement. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21338v1 Announce Type: new Abstract: The Model Context Protocol (MCP) has rapidly become the de facto standard for connecting large language models (LLMs) to external resources, but it also introduces a class of privacy risks that existing tools are ill-equipped to detect. Unlike conventional exfiltration bugs, leakage in MCP servers is largely protocol-induced: credentials, API keys, and Personally Identifiable Information (PII) cross the local/LLM boundary simply by being returned, logged, or raised inside a tool handler, with no explicit outbound request in the source code. We present MCPPrivacyDetector, a context-aware cross-language static analysis framework that detects such leakage in multilingual MCP servers. MCPPrivacyDetector lifts heterogeneous code implemented across different programming language (e.g., Python) into a unified program representation, applies context-aware semantic filtering to isolate genuinely sensitive values and protocol-specific implicit sinks (e.g., @mcp.tool handlers), and performs taint analysis to enumerate feasible flows. Applied to 10,655 real-world MCP servers, MCPPrivacyDetector finds leakage rates above 10%. Case studies confirm concrete exposures including leaked Bearer tokens, propagated API keys, and plaintext authentication credentials, arguing for systematic, protocol-aware safeguards in the emerging LLM agent toolchain.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21129unread
AgenticOS: An Intent-Oriented Secure Operating System Architecture for Autonomous AI Agents
Zhen Zhao, Yu Zhang, Yanpeng Zhu, Jia Wang, Songqiao Tao, Xin Cheng, Jiexin Gao · 2026-06-23
arXiv:2606. 21129v1 Announce Type: new Abstract: Traditional OS security models based on "resource exposure plus permission checks" face structural challenges as LLM-driven autonomous agents acquire capabilities for planning, tool use, network access, and code execution.
Read next because AgenticOS: An Intent-Oriented Secure Operating System Architecture for Autonomous AI 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: code, rect, soft, source, implement, capability, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21129v1 Announce Type: new Abstract: Traditional OS security models based on "resource exposure plus permission checks" face structural challenges as LLM-driven autonomous agents acquire capabilities for planning, tool use, network access, and code execution. Once an agent runtime is compromised through prompt injection or malicious tool outputs, an attacker can compose POSIX-style resource primitives into behaviors far beyond the user's task authorization. To address this, we propose AgenticOS, an intent-oriented secure OS architecture that consolidates delegable, auditable software capabilities into OS-native ones rather than replacing all applications. The core insight is to reframe the OS from a "resource manager" into an "intent filter": instead of requesting low-level resources directly, agents submit structured intent declarations, from which the system synthesizes a least-privilege environment with mandatory mediation, auditing, and information-flow constraints. At the implementation level, we introduce a four-layer architecture -- Ghost Kernel, Logic Shutter, Agent Capsule, and Semantic Boundary Gateway -- together with the Intent ABI, Manifest-Only Runtime, Weaver-based capability generation, and an admission model for AgenticOS-native Skills.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21067unread
Snatcher: Apple Find My Network Exposes Your Lost Devices To Strangers
Zhenyu Ren, Yanbo Zhang, Boya Liu, Mo Li · 2026-06-23
arXiv:2606. 21067v1 Announce Type: new Abstract: Apple's Find My network connects nearly one billion devices to locate missing property via Bluetooth Low Energy (BLE).
Read next because Snatcher: Apple Find My Network Exposes Your Lost Devices To Strangers 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, implement, without, full, test. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21067v1 Announce Type: new Abstract: Apple's Find My network connects nearly one billion devices to locate missing property via Bluetooth Low Energy (BLE). This paper reveals that insecure BLE advertisements and design tradeoffs allow unauthorized discovery and physical theft of lost Apple devices. We develop Snatcher, an attack and analysis framework implemented fully on Android smartphones without specialized hardware. Snatcher identifies vulnerabilities in unencrypted BLE advertisements, unauthenticated acoustic triggers, and slow MAC address randomization. Through three levels - sound-based direction finding, RSSI-IMU sensor-fusion navigation, and spatial-temporal clustering - our Android-based platform physically tracks and locates lost Apple accessories and devices in real-world tests. Our results highlight a crucial conflict between privacy protection, anti-stalking design, and physical security, urging Apple to strengthen Find My defenses.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21060unread
A Blockchain Consensus Mechanism for Distributed Electricity Trading
Shanglin Yang, Guo Chen, Shiping Chen · 2026-06-23
arXiv:2606. 21060v1 Announce Type: new Abstract: Distributed power systems complement centralized grids by coordinating distributed energy resources (DERs) to achieve regional energy self-sufficiency.
Read next because A Blockchain Consensus Mechanism for Distributed Electricity Trading 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, without, chain, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21060v1 Announce Type: new Abstract: Distributed power systems complement centralized grids by coordinating distributed energy resources (DERs) to achieve regional energy self-sufficiency. Scaling such systems raises four persistent challenges: decentralized coordination, fair economic settlement, trustworthy operation, and system optimization, all without a central authority. This paper proposes Proof of Energy (PoE), a blockchain consensus mechanism that addresses these challenges through cryptographically secured, contribution-proportional node selection. In PoE, block generation rights are tied directly to real-world energy contributions, enabling distributed consensus without centralized dispatch. An Energy Contribution Unit (ECU) model is introduced to map heterogeneous energy services onto a unified value metric via scarcity-weighted normalization. A Verifiable Random Function (VRF)-based proposal mechanism then ensures selection probability is strictly proportional to node contribution, preserving fairness and resisting manipulation. Case studies validate PoE across three dimensions: grid coordination, incentive fairness, and optimization efficiency. The result is a cryptographically secured, incentive-compatible framework for decentralized value distribution in energy systems.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21045unread
OVIG: Optimistic Verification of AI Training Integrity via Gradient Signals
Hongxu Su, Jianzhu Yao, Huan Zhang, Xuechao Wang, Pramod Viswanath · 2026-06-23
arXiv:2606. 21045v1 Announce Type: new Abstract: The rapid growth of AI has increased the demand for domain-specific post-training, while the cost and specialization of accelerator infrastructure push many model owners to outsource this process.
Read next because OVIG: Optimistic Verification of AI Training Integrity via Gradient Signals 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, under, source, rate, without, full, chain, language. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21045v1 Announce Type: new Abstract: The rapid growth of AI has increased the demand for domain-specific post-training, while the cost and specialization of accelerator infrastructure push many model owners to outsource this process. Outsourced training lowers operational barriers, but creates a training-integrity gap: the owner receives a checkpoint, logs, and aggregate metrics without direct evidence that the declared training trajectory was faithfully executed. An untrusted provider may have incentives to deviate from that trajectory, either to save computation or to introduce targeted security risks. Auditing such deviations is difficult because floating-point execution on heterogeneous accelerators introduces benign numerical drift, making it hard to distinguish honest replay differences from integrity violations. Existing verification methods either observe training at too coarse a granularity or impose costs and deployment constraints that are impractical at scale. We present OVIG, an optimistic verification framework that audits outsourced post-training using an empirical boundary on gradient differences calibrated from honest heterogeneous replays. OVIG checks opened intervals against this boundary and combines optimistic sampling with a stride parameter $s$, which partitions training into stride-aligned intervals and retains only interval-endpoint evidence. Across shortcut training attacks and targeted manipulation attacks, OVIG maintains $0\%$ ASR on language, vision, and diffusion workloads. On Qwen3, increasing the stride from $s=1$ to $s=2000$ reduces off-chain storage and evidence transmission by $1996\times$ while preserving $0\%$ ASR; at this setting, OVIG incurs only $1.143\times$ total system overhead relative to training without verification. These results show that OVIG provides a practical integrity layer for outsourced AI post-training under heterogeneous execution.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.20910unread
Whose Agent Are You? Multi-Layer Fingerprinting and Attribution of Autonomous Web Agents
Dayeon Kang, Hyejun Jeong, Jade Sheffey, Pubali Datta, Amir Houmansadr · 2026-06-23
arXiv:2606. 20910v1 Announce Type: new Abstract: As AI web agents proliferate, combining large language models with autonomous, browser-level control, indiscriminate content scraping by web agents has emerged as a privacy and security challenge.
Read next because Whose Agent Are You? Multi-Layer Fingerprinting and Attribution of Autonomous Web 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 "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, latin, line, rate, implement, control, full, language. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.20910v1 Announce Type: new Abstract: As AI web agents proliferate, combining large language models with autonomous, browser-level control, indiscriminate content scraping by web agents has emerged as a privacy and security challenge. Existing defenses, such as robots.txt and active bot-blocking, are insufficient, as they are widely violated and easily circumvented. In this work, we demonstrate that AI web agents can be effectively distinguished from humans and traditional crawlers using a multi-layer fingerprint based on both network layer characteristics (e.g., TLS, HTTP) and browser interaction behavior. We implement this mechanism as a programmatic logging framework that can be deployed on a live, instrumented domain. By analyzing six prominent agent frameworks (AutoGen, Browser Use, Claude, Gemini, Operator, and Skyvern), we uncover latent structural differences in how these systems assemble HTTP requests, establish TLS/HTTP connections, and execute autonomous browser actions. Feeding these multi-layer features into a decision tree classifier, our framework achieves high-fidelity identification (97% accuracy), successfully isolating distinct agent architectures and differentiating agent traffic from both human browsing baselines and legacy crawlers. Our findings demonstrate that cross-layer agent tracking provides a robust, evasion-resistant strategy for content protection and web security policy enforcement.
- score 94arxiv cs.LG (Machine Learning)arxiv:2606.21072unread
An Efficient and Effective Architecture for Large-Scale Traffic Prediction via Geometry-Adaptive Square Partitioning
Yongfeng Su, Hongwen Li, Zijian Zhang, Ziquan Fang, Lu Chen, Christian S. Jensen · 2026-06-23
arXiv:2606. 21072v1 Announce Type: new Abstract: Traffic prediction is a core task in intelligent transportation systems and urban-scale decision making.
Read next because An Efficient and Effective Architecture for Large-Scale Traffic Prediction via Geometry-Adaptive Square Partitioning 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, line, rate, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.21072v1 Announce Type: new Abstract: Traffic prediction is a core task in intelligent transportation systems and urban-scale decision making. Despite the effectiveness of mainstream neural-network based methods, their deployment in real-world settings with thousands of traffic sensors is jeopardized severely by their poor computational scalability. To address this, the community has attempted to incorporate spatial database partitioning techniques (e.g., Grid, Quadtree, and K-D Tree) to improve model scalability. However, these approaches rely on handcrafted geometric heuristics and often produce irregular or imbalanced data partitions, leading to boundary fragmentation, excessive padding overheads, and degraded model accuracy. In this paper, we propose SqLinear, an efficient and effective architecture for large-scale traffic prediction. First, we design Square Partition, a geometry-adaptive algorithm that partitions massive traffic sensors into balanced, non-overlapping, and near-square spatial regions. Unlike existing heuristic-based designs, Square Partition is theoretically grounded and provides provable guarantees on aspect ratio, balance, and partition utilization, establishing a high-quality foundation for downstream spatiotemporal modeling. Next, we propose a Hierarchical Linear Interaction (HLI) module that abandons the costly attention mechanisms commonly used in Transformer-based spatio-temporal models. HLI efficiently captures both local intra-region dynamics and global inter-region dependencies through a lightweight linear interaction scheme, enabling effective spatiotemporal modeling with linear computational complexity. Extensive experiments on four large-scale traffic datasets and 10 baselines show that SqLinear reduces MAE by 2.30% on average under standard setting and by 5.81% under extreme scalability settings, while reducing training runtime by 13.27%--30.84% in spatial- and horizon-scaling scenarios.
- score 94arxiv stat.ML (Machine Learning)arxiv:2606.21683unread
Finite-Sample Performance of Gradient Descent in Logistic Regression with Gaussian Design
Junren Chen, Arya Mazumdar · 2026-06-23
arXiv:2606. 21683v1 Announce Type: new Abstract: We consider the parameter estimation problem in logistic regression with Gaussian design: the estimation of a fixed unknown parameter $\theta^*\in \mathbb{R}^d$ ($\|\theta^*\|_2\ge 1$) from $n$ i.
Read next because Finite-Sample Performance of Gradient Descent in Logistic Regression with Gaussian Design 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, rate, control. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.21683v1 Announce Type: new Abstract: We consider the parameter estimation problem in logistic regression with Gaussian design: the estimation of a fixed unknown parameter $\theta^*\in \mathbb{R}^d$ ($\|\theta^*\|_2\ge 1$) from $n$ i.i.d. samples $\{(x_i,y_i)\}_{i=1}^n$, where $x_i\sim N(0,I_d)$ and $y_i|x_i \sim {\rm Bernoulli}(1/(1+\exp(-x_i^\top \theta^*)))$. Our main aim is to characterize the finite-sample estimation performance and convergence behavior of gradient descent (GD) on the maximum likelihood objective (i.e., the logistic loss). Under small $O(1)$ stepsize and $0$ initialization, we show that GD linearly converges to a small neighborhood of $\theta^*$ achieving an $\ell_2$ error of order $O(\sqrt{\|\theta^*\|_2^5d/n})$. This substantially goes beyond existing theoretical results that lack non-asymptotic estimation error rate and exhibit much slower parameter convergence. We also establish a faster local linear convergence to the same statistical error under a large $\Theta(\|\theta^*\|_2)$ stepsize. The main technical component is to show that the gradient of the logistic loss satisfies a certain approximate invertibility condition (AIC). To that end, we uniformly control the deviation of the gradient from its population counterpart by covering and peeling arguments, and then show that the population GD is a contraction by a delicate analysis based on the eigenvalues of population Hessian matrices. Finally, we build upon the recent work Matsumoto and Mazumdar (2025) and devise a novel efficient estimator that attains a sharper rate in high dimensions. This indicates that the existing non-asymptotic guarantees exhibit sub-optimal dependence on $\|\theta^*\|_2$, and that in many regimes $\Theta(\sqrt{\|\theta^*\|_2d/n})$ is the tight estimation error rate. Numerical examples are provided to corroborate our theoretical results.
- score 90arxiv stat.ML (Machine Learning)arxiv:2606.23662unread
Action-BED: Task-Driven Bayesian Experimental Design with Singly Intractable Objectives
Tom Rossa, Angus Phillips, Tom Rainforth · 2026-06-23
arXiv:2606. 23662v1 Announce Type: new Abstract: Bayesian experimental design (BED) has traditionally been based on maximising expected uncertainty reductions from prior to posterior.
Read next because Action-BED: Task-Driven Bayesian Experimental Design with Singly Intractable Objectives 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 "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation". Matching terms: eval, rate, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.23662v1 Announce Type: new Abstract: Bayesian experimental design (BED) has traditionally been based on maximising expected uncertainty reductions from prior to posterior. A major shortfall of this approach is that it leads to doubly intractable objectives that are difficult to optimise, while customising them to particular downstream tasks of interest can also be difficult. Following first principles decision theory, we demonstrate that BED can alternatively be formulated in terms of an expected future loss (EFL) on downstream actions, providing a simple and naturally task-driven framework. Critically, we then show that all such EFLs can be rearranged into singly intractable objectives that can be jointly optimised with respect to both the design policy and a downstream action policy using stochastic gradients, an approach we refer to as ACTION-BED. This formulation further sidesteps the need for any explicit posterior or marginal likelihood estimation and is naturally implicit, requiring only the ability to sample from the joint model over model parameters and data, and evaluate the downstream loss function. It thus allows design policies to be learned more effectively, efficiently, and simply than existing methods, while providing easy customisation to different downstream tasks and losses.
- score 86arxiv cs.CR (Cryptography and Security)arxiv:2606.21389unread
From Production SIEM to Reusable Cybersecurity Artifacts
Sidnei Barbieri, Leonardo Vaz de Meneses, \'Agney Lopes Roth Ferraz, Wagner Comin Sonaglio, Louren\c{c}o Alves Pereira J\'unior · 2026-06-23
arXiv:2606. 21389v1 Announce Type: new Abstract: Operational evidence is not automatically scientific evidence.
Read next because From Production SIEM to Reusable Cybersecurity Artifacts 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 "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation", 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, rate, language, model, absent. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21389v1 Announce Type: new Abstract: Operational evidence is not automatically scientific evidence. The most realistic Security Operations Center (SOC) data is production telemetry, yet it remains scientifically inaccessible because raw logs cannot be released; as a result, research relies on synthetic or dated datasets. We treat the boundary between private production telemetry and reusable research artifacts as the design object: a methodology that extracts, anonymizes, structures, and validates Security Information and Event Management (SIEM) data from a production financial SOC while preserving task-relevant investigative structure within a declared privacy boundary. Two consumers stress the same artifact. As training material, it fails loudly: 37 MITRE ATT&CK-mapped HIKARI challenges work only when anonymization preserves temporal order and entity consistency. As a measurement substrate, it fails quietly: across 200 SOCpilot incidents, a deterministic verifier detects non-compliant Large Language Model (LLM) actions that are absent from the human baseline. The result is a measurable privacy-utility boundary rather than a formal anonymity claim.
- score 86arxiv cs.CR (Cryptography and Security)arxiv:2606.21377unread
ARENA: An Architecture for Measuring the Transferability of Autonomous Cyber Defense
Sidnei Barbieri, \'Agney Lopes Roth Ferraz, Wagner Comin Sonaglio, Gioliano de Oliveira Braga, Henrique Curi de Miranda, Louren\c{c}o Alves Pereira J\'unior · 2026-06-23
arXiv:2606. 21377v1 Announce Type: new Abstract: Operational evidence is not automatically scientific evidence.
Read next because ARENA: An Architecture for Measuring the Transferability of Autonomous Cyber Defense 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 "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation", 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, rate, language, model, absent. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21377v1 Announce Type: new Abstract: Operational evidence is not automatically scientific evidence. The most realistic Security Operations Center (SOC) data is production telemetry, yet it remains scientifically inaccessible because raw logs cannot be released; as a result, research relies on synthetic or dated datasets. We treat the boundary between private production telemetry and reusable research artifacts as the design object: a methodology that extracts, anonymizes, structures, and validates Security Information and Event Management (SIEM) data from a production financial SOC while preserving task-relevant investigative structure within a declared privacy boundary. Two consumers stress the same artifact. As training material, it fails loudly: 37 MITRE ATT&CK-mapped HIKARI challenges work only when anonymization preserves temporal order and entity consistency. As a measurement substrate, it fails quietly: across 200 SOCpilot incidents, a deterministic verifier detects non-compliant Large Language Model (LLM) actions that are absent from the human baseline. The result is a measurable privacy-utility boundary rather than a formal anonymity claim.
- score 78arxiv cs.AI (Artificial Intelligence)arxiv:2606.20637unread
Constituency Optimisation Through Hamiltonian Representation Of Mandates (COTHROM): Algorithmic Redistricting of Irish Election Boundaries
Ruaidhr\'i Campion, Matthew Fenlon, Joshua Cooney Mercedal, Casey Farren-Colloty, Eliza Somerville, Michael A. J. Mitchell · 2026-06-23
arXiv:2606. 20637v1 Announce Type: new Abstract: Electoral redistricting in Ireland's Proportional Representation Single Transferable Vote (PR-STV) system faces the challenge of selecting an optimally representative set of electoral boundaries from an enormous set of possible configurations, and where ``representative'' is a delicate balance of constitutional objectives that are often in tension with one another.
Read next because Constituency Optimisation Through Hamiltonian Representation Of Mandates (COTHROM): Algorithmic Redistricting of Irish Election Boundaries 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)", 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 "Can capability be taught through another persona?". Matching terms: eval, chain, another. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20637v1 Announce Type: new Abstract: Electoral redistricting in Ireland's Proportional Representation Single Transferable Vote (PR-STV) system faces the challenge of selecting an optimally representative set of electoral boundaries from an enormous set of possible configurations, and where ``representative'' is a delicate balance of constitutional objectives that are often in tension with one another. We present the first computational framework for Irish electoral redistricting that systematically optimises across multiple constitutional requirements while making trade-offs explicit and quantifiable. The electoral redistricting problem is parsed using statistical physics, where constitutional objectives are considered as terms in a Potts Hamiltonian. Markov Chain Monte Carlo (MCMC) methods and simulated annealing are employed to minimise this objective function, systematically exploring this configuration space, with coupling constants as proxies for objective weightings. Multi Criterion Decision Analysis (MCDA) and Pareto Optimality is then utilised to remedy the ambiguity in choosing a certain objective weighting combination over others. With respect to proportional representation and compactness objectives evaluated in County Cork, COTHROM consistently improves on the existing legal constituency boundaries for a range of objective weightings.
- score 62arxiv cs.AI (Artificial Intelligence)arxiv:2606.20656unread
Learning Splitting Heuristics for Parallel String Solvers
Chenhao Gao, Peisen Yao · 2026-06-23
arXiv:2606. 20656v1 Announce Type: new Abstract: String constraint solvers are crucial for reasoning about string-manipulating programs.
Read next because Learning Splitting Heuristics for Parallel String Solvers 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)", experiment "Implement Chen et al. persona-vector extraction recipe and compare to project's centroid-difference recipe". Matching terms: latin, implement. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20656v1 Announce Type: new Abstract: String constraint solvers are crucial for reasoning about string-manipulating programs. However, many practical string constraints are undecidable, and real-world applications often present complex constraints that challenge current solvers. The rise of multi-core architectures offers an opportunity for parallel solving. A key parallel solving method is \emph{cube-and-conquer}, in which the quality of splitting heuristics is critical to effectively dividing the search space. Unfortunately, manually designing the heuristics is labor-intensive, and handcrafted heuristics are often sub-optimal. This paper introduces a data-driven approach to automatically generating splitting heuristics. We frame the problem of selecting a splitting atom as a learning task, using features from input formulas and dynamic data from solver execution. We implement this approach in two popular string solvers, Z3seq and Z3str4, demonstrating that the learned heuristics outperform manually designed ones in the number of solved formulas and the average solving time.
- score 62arxiv cs.AI (Artificial Intelligence)arxiv:2606.20622unread
Darwin Mobile Agent: A Roadmap for Self-Evolution
Daniel Beechey, Derek Yuen, Jianheng Liu, Dezhao Luo, Tiantian He, Weilin Luo, Jun Wang, Kun Shao · 2026-06-23
arXiv:2606. 20622v1 Announce Type: new Abstract: The goal of artificial intelligence is to create agents capable of general, adaptive behaviour in open-ended environments.
Read next because Darwin Mobile Agent: A Roadmap for Self-Evolution overlaps with 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)", experiment "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation". Matching terms: source, stage. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20622v1 Announce Type: new Abstract: The goal of artificial intelligence is to create agents capable of general, adaptive behaviour in open-ended environments. Guided by the "Bitter Lesson", we argue that the most effective path toward this goal is to systematically remove human priors and allow intelligence to naturally emerge through interaction with a "Big World" that is orders of magnitude more complex than the agent itself. We propose the mobile Graphical User Interface (GUI) as a practical proxy for such a world and introduce Darwin Mobile Agent, an open-source infrastructure designed as a foundation for autonomous reinforcement learning in this domain. This framework addresses the data-collection bottleneck in real-world mobile interactions by using an asynchronous agent-environment loop across parallel cloud-phone instances. We further propose a conceptual roadmap to systematically remove human priors from three fundamental pillars of a self-evolving agent: task curricula, outcome verification, and memory management. We validate that the Darwin infrastructure provides the stability and scalability required for the first stage of this roadmap: policy optimisation in the GUI domain. This work establishes the practical and theoretical foundation necessary to move toward truly autonomous, self-evolving GUI agents.
- score 62arxiv stat.ML (Machine Learning)arxiv:2606.21199unread
Orthogonal Discrepancy Kernels for Learning with Partial Physics
Swapnil Manna, Timothy J. Rogers, Lawrence Bull · 2026-06-23
arXiv:2606. 21199v1 Announce Type: new Abstract: We introduce a semi-parametric framework for nonlinear system identification, which decouples discrepancy functions from physics-based components.
Read next because Orthogonal Discrepancy Kernels for Learning with Partial Physics 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, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.21199v1 Announce Type: new Abstract: We introduce a semi-parametric framework for nonlinear system identification, which decouples discrepancy functions from physics-based components. Orthogonal Gaussian process regression balances sparse parameter selection (the white box) with discrepancy learning (the black box) to produce interpretable models from incomplete physics.
- score 58arxiv stat.ML (Machine Learning)arxiv:2606.21260unread
Subsampling for supervised learning in reproducing kernel Hilbert spaces
Eyal Vayness, Maxime Sangnier · 2026-06-23
arXiv:2606. 21260v1 Announce Type: new Abstract: In the era of big data, subsampling became a common practice in statistical learning.
Read next because Subsampling for supervised learning in reproducing kernel Hilbert spaces overlaps with 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 "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: trained. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.21260v1 Announce Type: new Abstract: In the era of big data, subsampling became a common practice in statistical learning. By selecting a subgroup of individuals based on which the learner is trained, subsampling aims at reducing the computational cost and time of the estimation step, and ideally leads to a decrease of its energy consumption and carbon footprint. This work focuses on a nonparametric setting, in which the hypotheses set lies in a reproducing kernel Hilbert space, and the estimator is a minimizer of an empirical risk reweighted \`a la Horvitz-Thompson. By studying the asymptotic properties of this estimator, we reveal an optimal subsampling scheme (regarding the trace of the covariance operator) and show that it can be used via plug-in. A numerical study on synthetic and real-world datasets shows the practicability and the benefit of the proposed approach.
- score 46arxiv cs.AI (Artificial Intelligence)arxiv:2606.20643unread
SPARC: A Multi-Agent System for Electrical Circuit Question Answering
Mushtari Sadia, Zhenning Yang, Umme Habiba Lamia, Nishat Shawrin, Ang Chen, Amrita Roy Chowdhury · 2026-06-23
arXiv:2606. 20643v1 Announce Type: new Abstract: Electrical circuit diagram QA tasks require complex mathematical reasoning, which remains challenging for multimodal LLMs.
Read next because SPARC: A Multi-Agent System for Electrical Circuit Question Answering 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)". Matching terms: line. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20643v1 Announce Type: new Abstract: Electrical circuit diagram QA tasks require complex mathematical reasoning, which remains challenging for multimodal LLMs. We present SPARC, a multi-agent system that answers questions over circuit diagrams by grounding reasoning in executable physics-based simulations. SPARC uses LLM agents to synthesize, execute, and analyze simulation programs, improving accuracy and reliability by design. It achieves 83% accuracy, with up to a 58% absolute improvement over baselines, while enabling systematic error diagnosis.
- score 46arxiv stat.ML (Machine Learning)arxiv:2606.21185unread
Two Layers of Instability in Causal Estimation
Alexis Bellot · 2026-06-23
arXiv:2606. 21185v1 Announce Type: new Abstract: There is a precise sense in which drawing causal inferences from observational data is hard, even when identifiability is assumed.
Read next because Two Layers of Instability in Causal Estimation overlaps with 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: model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.21185v1 Announce Type: new Abstract: There is a precise sense in which drawing causal inferences from observational data is hard, even when identifiability is assumed. In particular, Robins and Ritov (1997) and Robins et al. (2003) showed that causal effects can be discontinuous as a function of the data distribution: two arbitrarily close data distributions might correspond to different causal effects. This is a fact independent of the choice of estimator; however, not all estimators are equally unstable. Our contribution is to surface a second layer of instability that depends on the choice of estimator. We show that many standard point estimates can be read as point summaries of multimodal distributions over the space of structural causal models. As such, estimators can jump discontinuously in the data distribution. This defines a taxonomy of estimators that admits a decision-theoretic reading: stability depends on whether the implicit loss function an estimator optimizes is aligned with the causal effect itself. Specifically, inverse propensity weighted estimators and regression estimators are examples of discontinuous summaries, while explicit posterior means and medians are shown to be continuous.
Threats and caveats
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20677unread
Democratizing and accelerating AI-driven pathology research through agentic intelligence
Jiabo Ma, Cheng Jin, Yihui Wang, Hao Jiang, Ling Liang, Yingxue Xu, Junlin Hou, Zhengrui Guo, Zhengyu Zhang, Yifei Xia, Hongyi Wang, Fengtao Zhou, Zhe Xu, Huajun Zhou, Jiarui Ouyang, Qian Zeng, On Ki Tang, Eunhyang Park, Carolyn Glass, Ronald Cheong Kin Chan, Li Liang, Hao Chen · 2026-06-23
arXiv:2606. 20677v1 Announce Type: new Abstract: Computational pathology has advanced rapidly with the emergence of foundation models, yet widespread adoption remains limited by substantial technical complexity and programming requirements.
Read next because Democratizing and accelerating AI-driven pathology research through agentic intelligence 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, rate, implement, control, without, position. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20677v1 Announce Type: new Abstract: Computational pathology has advanced rapidly with the emergence of foundation models, yet widespread adoption remains limited by substantial technical complexity and programming requirements. Here we present PathLab, an autonomous agentic framework that translates natural-language research objectives into executable and validated computational pathology workflows through the structured composition of domain-specific skills and tools. By organizing workflow generation around reusable methodological modules, including data preprocessing, model development, evaluation and interpretation, PathLab enables studies to be specified at the level of scientific intent rather than implementation details. We evaluated PathLab across 12 public datasets spanning four representative task families: region-of-interest classification, whole-slide image classification, segmentation and survival prediction. Across all task categories, PathLab achieved non-inferior performance relative to expert implementations, while consistently enforcing semantic validity of user prompts and proactively rejecting incompatible workflow specifications prior to execution. In controlled user studies, PathLab substantially reduced the time required to generate executable analytical pipelines and enabled domain experts without programming experience to independently design, execute and evaluate computational pathology studies. Together, these results establish PathLab as a reliable interface between biomedical intent and computational execution, enabling computational pathology studies to be designed at the level of scientific questions rather than programming expertise. By lowering technical barriers to advanced AI methodologies, PathLab provides a foundation for the broader democratization of computational pathology.
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 cs.AI (Artificial Intelligence)arxiv:2606.20667unread
A Quantum-Assisted Agentic Distributed Artificial Intelligence Framework for Deadline-Bounded Orchestration of Hybrid Renewable Microgrids
Iacovos I. Ioannou, Saher Javaid, Minella Bezha, Yasuo Tan, Naoto Nagaoka, Vasos Vassiliou · 2026-06-23
arXiv:2606. 20667v1 Announce Type: new Abstract: The real-time orchestration of microgrids that combine fluctuating renewable sources, dispatchable units, storage and curtailable consumers requires the repeated solution of combinatorial dispatch and coalition formation problems under hard control deadlines.
Read next because A Quantum-Assisted Agentic Distributed Artificial Intelligence Framework for Deadline-Bounded Orchestration of Hybrid Renewable Microgrids 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, eval, source, line, control, trained. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20667v1 Announce Type: new Abstract: The real-time orchestration of microgrids that combine fluctuating renewable sources, dispatchable units, storage and curtailable consumers requires the repeated solution of combinatorial dispatch and coalition formation problems under hard control deadlines. In this paper, a quantum-assisted agentic distributed artificial intelligence (DAI) framework is proposed in which the dispatch problem of each control slot is formulated as a quadratic unconstrained binary optimization (QUBO) problem by Belief-Desire-Intention extended (BDIx) agents and is solved by a portfolio of quantum, quantum-inspired and classical solvers. Solver selection is elevated to a first-class agentic deliberation action of the coordinator agent. Learned beliefs about solver latencies are maintained and the solver intention that is expected to satisfy the prevailing deliberation deadline is committed in each slot. In addition, a belief-shaped storage valuation mechanism is introduced through which the storage agent prices its energy at a discounted future-peak value, injecting intertemporal information into the otherwise myopic per-slot optimization. The framework is evaluated on a 24-hour simulation of a grid-connected microgrid with photovoltaic, wind, battery, genset and demand-response assets, with the Quantum Approximate Optimization Algorithm (QAOA) executed by statevector simulation and benchmarked per slot against tabu search, simulated annealing, binary particle swarm optimization, greedy descent and exhaustive enumeration. Zero deliberation deadlines are missed, the committed dispatch attains the exact optimum on every slot and the realized daily cost of 146.24 EUR equals the exact lower bound, with 97.83 percent renewable utilization and zero unserved energy. When the storage valuation mechanism is deactivated, the daily cost is increased to 152.75 EUR, a 4.5 percent increase.
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:2606.20663unread
DrugBench: Evaluating AI Control Protocols for Medication Harm Mitigation
Guido Freire, Agust\'in Mart\'inez-Su\~n\'e, Viviana Cotik · 2026-06-23
arXiv:2606. 20663v1 Announce Type: new Abstract: Large Language Models have the potential to expand and improve the access to clinical information by enabling new ways of interacting with medical knowledge in natural language.
Read next because DrugBench: Evaluating AI Control Protocols for Medication Harm Mitigation 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, line, control, language, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20663v1 Announce Type: new Abstract: Large Language Models have the potential to expand and improve the access to clinical information by enabling new ways of interacting with medical knowledge in natural language. However, their deployment in medical question-answering settings is safety-critical, since misaligned outputs can lead to severe patient harm. AI control is an emerging approach that introduces external safeguards to mitigate unsafe behaviours in misaligned systems and has been shown to be effective in domains such as code generation. However, its applicability and effectiveness in medical settings have not been systematically studied. In this work, we present a pipeline for evaluating AI control protocols to mitigate medication-related harm. To this end, we introduce DrugBench, an AI control evaluation benchmark which combines 3,671 multi-turn medical conversations from HealthBench with drug information from official FDA labels, covering four categories of medication-related harm: drug interactions, contraindications, dosing constraints, and patient action restrictions. Furthermore, inspired by the medical domain, we argue that safety should account for the severity of unsafe outputs, not just their probability. Under this revised definition, we show that existing control protocols can be subverted and propose severity-based monitoring to address this limitation.
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, evaluation, benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20662unread
Confidence Laundering in Agent Systems: Why Uncertainty Needs a Latent Carrier
Kaiwen Shi, Zheyuan Zhang, Han Bao, Colby Nelson, Yanfang Ye · 2026-06-23
arXiv:2606. 20662v1 Announce Type: new Abstract: Modern agent systems can turn uncertainty into overconfidence.
Read next because Confidence Laundering in Agent Systems: Why Uncertainty Needs a Latent Carrier 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, does, propagate, position. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20662v1 Announce Type: new Abstract: Modern agent systems can turn uncertainty into overconfidence. Fragile upstream decisions are often exposed to downstream components as clean intermediate artifacts, while the uncertainty behind those decisions is lost at the interface. As a result, local ambiguity can become system-level error amplification. We argue that this reveals an interface bottleneck in agent uncertainty propagation: uncertainty does not propagate simply because a trajectory contains uncertain steps; it propagates only when it survives the handoff between components. We define uncertain decision handoff as the transfer of an intermediate decision made under uncertainty, and identify confidence laundering as a failure mode in which fragile upstream states are repackaged as procedurally valid artifacts that downstream agents over-trust. To address this bottleneck, we propose latent uncertainty as an uncertainty-bearing carrier attached to decision handoffs. Rather than replacing text with hidden states, latent uncertainty aims to preserve pre-commitment fragility in a form that downstream components can use. This position shifts agent uncertainty propagation from step-wise uncertainty estimation toward uncertainty-preserving interface design for more recoverable agent systems.
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:2606.20661unread
From Knowing to Acting: Benchmarking Self-Awareness Capability of LLM Agents
Yifan Li, Shengbin Yue, Boyu Feng, Jinhu Qi, Bo Ke, Zixing Song, Hongru Wang, Zhongyu Wei, Irwin King · 2026-06-23
arXiv:2606. 20661v1 Announce Type: new Abstract: The integration of external tools has transitioned LLM agents from passive responders to autonomous systems.
Read next because From Knowing to Acting: Benchmarking Self-Awareness Capability of LLM 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: code, strong, alignment, eval, source, rate, capability, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20661v1 Announce Type: new Abstract: The integration of external tools has transitioned LLM agents from passive responders to autonomous systems. However, current benchmarks prioritize execution success, neglecting self-awareness capability, the ability to discern whether a problem requires necessary external resources or can be solved via internal parametric knowledge. To address this, we introduce KAPRO (Knowing-Acting Quadrant PRObe), a framework that evaluates cognitive-behavioral alignment by decoupling an agent's metacognitive judgment (Knowing) from its spontaneous execution (Acting). We further construct KAware, a dataset rigorously partitioning tasks into external, internal, and hybrid subspaces to systematically probe these epistemic boundaries. Extensive experiments across diverse agent architectures show that self-awareness capability is strongly correlated with task success but degrades sharply in internal-capability settings. Moreover, open-source and instruction-following models exhibit stronger tool overuse due to shallow pattern matching, while proprietary and reasoning-oriented models demonstrate more reliable cognitive gating. Benchmark and codes are available at https://github.com/AI-Santiago/KAware.
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:2606.20659unread
Skill Coverage: A Test Adequacy Metric for Agent Skills
Boyin Tan, Xiaowei Huang, Youcheng Sun · 2026-06-23
arXiv:2606. 20659v1 Announce Type: new Abstract: Agent skills encode reusable procedural knowledge that guides large language model agents across tasks and execution contexts.
Read next because Skill Coverage: A Test Adequacy Metric for Agent Skills 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, without, alone, does, contexts. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20659v1 Announce Type: new Abstract: Agent skills encode reusable procedural knowledge that guides large language model agents across tasks and execution contexts. Existing evaluations primarily assess skills through task level outcomes, yet task success alone does not reveal which parts of a skill have been exercised or which remain untested. We introduce skill coverage, a test adequacy metric that treats the skill artifact as the object under test. Our approach extracts observable skill behavior constraints from skill documents and measures whether an agent trajectory provides sufficient evidence to exercise and evaluate each constraint. Skill coverage uses a binary cover or not cover judgment, which reports whether a documented behavior has been exercised with sufficient observable evidence, without assigning an additional outcome label to the behavior. Applying skill coverage to SkillsBench reveals that existing benchmark executions cover only 39.90 to 43.98% of skill behavior constraints, suggesting that current benchmark tasks leave large portions of documented skill guidance untested. These findings show that successful task completion does not imply adequate testing of the skill artifact, highlighting skill coverage as a measure of how thoroughly agent skills are tested.
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:2606.20644unread
Bridging Multi-Valued Heuristics and Dimensionality Reduction in Multi-Objective Search
Maya Wolff, Ariel Felner, Oren Salzman · 2026-06-23
arXiv:2606. 20644v1 Announce Type: new Abstract: Multi-objective shortest-path (MOSP) algorithms traditionally rely on single-valued heuristics (SVHs), which associate each state with a single admissible cost vector.
Read next because Bridging Multi-Valued Heuristics and Dimensionality Reduction in Multi-Objective Search 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, test. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20644v1 Announce Type: new Abstract: Multi-objective shortest-path (MOSP) algorithms traditionally rely on single-valued heuristics (SVHs), which associate each state with a single admissible cost vector. While SVHs provide safe lower bounds, they fail to capture the trade-off structure of the Pareto frontier and often yield weak search guidance. Multi-valued heuristics (MVHs) address this limitation by mapping states to sets of cost estimates, enabling a richer approximation of possible trade-offs. Modern MOSP algorithms are highly dependent on dimensionality reduction (DR) techniques to efficiently perform dominance checks. However, integrating MVHs with DR introduces subtle correctness challenges. We show that naively combining DR with MVHs destroys the ordering invariants required for DR, leading to unsound and incomplete search. To address this issue, we develop the first theoretical frameworks for safely integrating MVHs with DR. First, we introduce $\text{NAMOA}^*{\text{dr}\text{-}\text{mvh}}$, a theoretical baseline that restores search correctness by enforcing heuristic consistency. Recognizing the practical limitations of this approach, we then introduce our primary contribution, $\text{L}\text{-}\text{NAMOA}^*{\text{dr}\text{-}\text{mvh}}$. This algorithm employs a "lazy," optimistic approach to DR, preserving exact correctness with only an admissible MVH by dynamically detecting and repairing local ordering violations. Across a range of benchmarks, $\text{L}\text{-}\text{NAMOA}^*{\text{dr}\text{-}\text{mvh}}$ matches or improves over state-of-the-art MOSP algorithms, and achieves speedups of over 10x in instances where the additional guidance provided by the MVH translates into stronger pruning.
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, benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20638unread
RIZZ: Routing Interactions to Near Zero-Interference Zones for Continual Adaptation of Black-Box Agents
Sonali Goel, Pranav Vaidhyanathan, Lucas Schorling, Natalia Ares, Maike Osborne · 2026-06-23
arXiv:2606. 20638v1 Announce Type: new Abstract: Large language models are increasingly deployed as long-lived agents that must adapt across users, tasks, domains, modalities, and feedback regimes without access to model weights.
Read next because RIZZ: Routing Interactions to Near Zero-Interference Zones for Continual Adaptation of Black-Box 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, line, rate, control, without, another, language. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20638v1 Announce Type: new Abstract: Large language models are increasingly deployed as long-lived agents that must adapt across users, tasks, domains, modalities, and feedback regimes without access to model weights. Existing black-box adaptation methods typically optimize a single prompt, maintain an undifferentiated memory, or rely on repeated rollout-heavy search. However, these designs struggle when streams of input are nonstationary, feedback is sparse, and failures from one task family can contaminate behavior on another. We introduce RIZZ (Routing Interactions to Near Zero-interference Zones), a continual adaptation framework for compound language-model systems that learns entirely through verifier-gated memory, routing, and prompt compilation. RIZZ organizes input streams into dynamically spawned memory branches. At inference time, either while online or offline, a context-aware router selects or creates a branch that retrieves branch-local, global, graph-structured, and working-memory context, which is compiled into a bounded prompt together with retrieved task evidence. After the model acts, task verifiers score the output, and only verified interactions can update memory, promote reusable rules, demote harmful rules, or create anti-patterns. This yields a black-box agent that improves through persistent natural-language feedback while explicitly controlling interference. RIZZ targets the regime where adaptation must occur online under context budgets. Finally, we demonstrate the effectiveness of our framework against state-of-the-art baselines on competitive 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 failure, failures, benchmark.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20636unread
SkillHarness: Harnessing Safe Skills for Computer-Use Agents
Yurun Chen, Biao Yi, Keting Yin, Shengyu Zhang · 2026-06-23
arXiv:2606. 20636v1 Announce Type: new Abstract: Computer-Use Agents (CUAs) are increasingly deployed in dynamic interactive environments, creating a growing need for continual skill learning during interaction.
Read next because SkillHarness: Harnessing Safe Skills for Computer-Use 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, source, line, rate, trained, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20636v1 Announce Type: new Abstract: Computer-Use Agents (CUAs) are increasingly deployed in dynamic interactive environments, creating a growing need for continual skill learning during interaction. Recent approaches address this challenge by learning reusable skills from successful trajectories. However, these skill learning methods largely assume static and safe environments, overlooking risks from adversarial interactions (e.g., prompt injections) and environmental dynamics (e.g., pop-ups). In dynamic settings, such assumptions can lead to risky skill learning and brittle execution, undermining the reliability of CUAs. This raises the question: how can CUAs learn and use skills safely in dynamic environments? To address this problem, we propose SkillHarness, a framework for safe skill harnessing in dynamic environments. SkillHarness moves beyond static skill abstractions by modeling skill learning and utilization as a safety-constrained interaction process. Specifically, we introduce the skill boundary that leverages multi-source supervision signals to identify safe skills from interaction trajectories, and construct self-improving safety constraints throughout the skill lifecycle. In addition, SkillHarness introduces selective skill reuse, where tasks are guided to decompose according to context and completed through the selective activation of skill subsets. Our experiments demonstrate that SkillHarness significantly reduces the unsafe rate of learned skills by 57.1% and consistently improves execution stability under dynamic environmental changes, outperforming existing 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 adversarial.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20634unread
DEMM-Bench: A Cross-Regime Benchmark for Agent-Runtime Governance-Evidence Sufficiency
Oleg Solozobov · 2026-06-23
arXiv:2606. 20634v1 Announce Type: new Abstract: Agent-runtime systems emit traces, ledgers, provenance graphs, policy logs, delegation tokens, cache events, and tool-firewall records, but those containers do not necessarily answer governance questions about a specific decision.
Read next because DEMM-Bench: A Cross-Regime Benchmark for Agent-Runtime Governance-Evidence Sufficiency 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, source, token, line, rate, emit, candidate. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20634v1 Announce Type: new Abstract: Agent-runtime systems emit traces, ledgers, provenance graphs, policy logs, delegation tokens, cache events, and tool-firewall records, but those containers do not necessarily answer governance questions about a specific decision. DEMM-Bench is a cross-regime benchmark for agent-runtime governance-evidence sufficiency, grounded in the Decision Evidence Maturity Model (DEMM): it measures whether records across eight evidence regimes are sufficient to reconstruct decision-level properties rather than merely present. The benchmark normalizes the regimes through adapters, asks property questions over actor, authority, action, policy, decision basis, resource touch, lifecycle context, and verification strength, and applies eight deterministic degradation conditions. Across 64 manuscript cases, trace-present and schema-present baselines overclaim on 75% of cases, ledger-present overclaims on 50%, and the redacted property-level candidate scorer has zero overclaim with 56.25% mean Property Sufficiency Accuracy. The deposited package provides the 64-case dataset, construction-oracle labels, baselines, and adapters, supporting reproducible evaluation of decision-evidence maturity across heterogeneous agent-runtime evidence substrates.
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:2606.20627unread
Latent Goal Prediction from Language for Model-Based Planning
Samuel Barbeau, Simon Roy, Giovanni Beltrame, Christian Desrosiers, Nicolas Thome · 2026-06-23
arXiv:2606. 20627v1 Announce Type: new Abstract: Planning with world models is bottlenecked by compounding prediction errors and the difficulty of defining optimizable goals.
Read next because Latent Goal Prediction from Language for Model-Based Planning 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, soft, eval, line, control, language, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20627v1 Announce Type: new Abstract: Planning with world models is bottlenecked by compounding prediction errors and the difficulty of defining optimizable goals. Visual targets provide precise local gradients but poor distant guidance, while language is flexible yet limited by noisy cross-modal alignment or dependence on large generative models unsuited for the high-sampling nature of model-based planning. To address these challenges, we introduce Latent Goal Prediction from Language (LAGO), a framework that predicts both sequences of intermediate goal states from language instructions and action-conditioned rollouts, all within the same latent space. Rather than optimizing toward a single global objective, LAGO dynamically decomposes instructions into explicitly predicted, locally tractable latent subgoals. By updating these subgoals online and using a soft minimum trajectory cost during planning, LAGO enables an agent to follow coherent latent trajectories over long horizons. Evaluation across multiple environments planning horizons shows that LAGO avoids the sharp degradation of prior methods. By achieving robust and precise long-horizon planning purely from language, LAGO bridges the precision of visual goals with the flexibility of text-guided 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.
- score 100arxiv cs.AI (Artificial Intelligence)arxiv:2606.20625unread
AlphaMemo: Structured Search-Process Memory for Self-Evolving Alpha Mining Agents
Hang Yu, Zifan Zheng, Jeff Z. Pan, Tongliang Liu, Zhiyong Wang, Fengxiang He · 2026-06-23
arXiv:2606. 20625v1 Announce Type: new Abstract: LLM agents are promising for alpha mining via combining financial priors, symbolic reasoning, executable factor generation, and feedback-driven refinement.
Read next because AlphaMemo: Structured Search-Process Memory for Self-Evolving Alpha Mining 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: code, text, under, alpha, control, full, factor, contexts. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20625v1 Announce Type: new Abstract: LLM agents are promising for alpha mining via combining financial priors, symbolic reasoning, executable factor generation, and feedback-driven refinement. Yet, they face a combinatorial search space, noisy non-stationary feedback, redundant discoveries, and overfitting risks from naively reusing past successes. To address these challenges, we propose AlphaMemo, a self-evolving alpha mining agent with Structured Search-Process Memory. Rather than memorizing only final factors or full trajectories, AlphaMemo records reusable evidence about which edit motifs work or fail under specific parent-factor contexts. It extracts motifs from Abstract Syntax Tree (AST) differences, applies confidence-gated residual memory on top of a search-ledger prior, and uses asymmetric veto control to suppress high-confidence failure patterns. Experiments on CSI 500 and S\&P 500 show improved out-of-sample performance and fixed-budget discovery efficiency, with ablations validating the roles of residual learning, confidence gating, AST-diff motifs, and veto memory. Code is at https://github.com/jarrettyu/AlphaMemo.
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:2606.20615unread
Specifying AI-SDLC Processes: A Protocol Language for Human-Agent Boundaries
Ylli Prifti · 2026-06-23
arXiv:2606. 20615v1 Announce Type: new Abstract: AI agents now participate as first-class team members across the software development lifecycle, yet no specification language exists for expressing the human-agent responsibility boundaries, approval gates, and governance constraints this collaboration requires.
Read next because Specifying AI-SDLC Processes: A Protocol Language for Human-Agent Boundaries 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, soft, eval, token, rate, implement, control. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20615v1 Announce Type: new Abstract: AI agents now participate as first-class team members across the software development lifecycle, yet no specification language exists for expressing the human-agent responsibility boundaries, approval gates, and governance constraints this collaboration requires. Existing approaches encode process in agent prompts (subject to drift), target adjacent domains (workflow management, business processes), or address only fragments (access control, approval gates). We propose a domain-specific language for specifying AI-SDLC processes as protocols, with formal syntax, well-formedness conditions, operational semantics, and enforcement invariants. The language distinguishes policy (declared intent) from mechanism (structural enforcement), enabling implementations to bound process non-determinism through primitives such as validation tokens and capability boundaries. Three results follow. A failure rate analysis shows that structural enforcement bounds system failure rates at a weighted product of agent and validator rates, while behavioral compliance permits cumulative or near-saturating growth. The 2+N team pattern (two human-in-control roles plus N specialized agent members) formalizes classical Separation of Duties for AI-SDLC. Kleene closure of orchestration loops and reflexive protocol-adherence validation emerge as design properties rather than special-case constructs. We position the contribution against multi-agent frameworks (MetaGPT), workflow specification (FlowAgent, BPMN extensions), and capability-based security (SAGA): the novelty lies in the specific integration, not any single primitive. A working implementation demonstrates feasibility; empirical evaluation is future work.
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.AI (Artificial Intelligence)arxiv:2606.20599unread
Beyond Fixed Budgets: Characterizing the Inelasticity and Limitations of Tree-of-Thought Reasoning Strategies
Atkia Mahila, Avinash Maurya, M. Mustafa Rafique, Bogdan Nicolae · 2026-06-23
arXiv:2606. 20599v1 Announce Type: new Abstract: Tree of Thought (ToT) search has become a promising direction for improving the reasoning capabilities of large language models, but deploying these methods in practice raises a question that has received little systematic attention: how do different search strategies behave under varying compute budgets, model sizes, and problem difficulties?
Read next because Beyond Fixed Budgets: Characterizing the Inelasticity and Limitations of Tree-of-Thought Reasoning Strategies 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, eval, source, token, rate, trained. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20599v1 Announce Type: new Abstract: Tree of Thought (ToT) search has become a promising direction for improving the reasoning capabilities of large language models, but deploying these methods in practice raises a question that has received little systematic attention: how do different search strategies behave under varying compute budgets, model sizes, and problem difficulties? In this work, we evaluate two representative ToT methods; DPTS, a Monte Carlo tree search based approach, and SSDP, a semantic deduplication based approach, across two mathematical reasoning benchmarks (Math500 and GSM8K), two model scales (Llama-3B and Llama-8B), and four token budgets (3k--10k). Our analysis reveals that the two methods exhibit limitations that pull in opposite directions. DPTS suffers from a cold-start bottleneck at low budgets: it requires sufficient exploration before its value estimates become reliable, making it a poor fit for resource-constrained settings despite strong scaling behavior at higher budgets. SSDP, on the other hand, reaches candidate solutions efficiently but is prone to frontier depletion; its aggressive node merging permanently discards unexplored paths, leaving it unable to improve regardless of how much budget remains. Together, these findings suggest that neither a fixed exploration strategy nor a fixed pruning strategy is sufficient across compute continuum. We argue that effective search for scientific reasoning agents requires strategies that can adapt their behavior based on search progress and available resources.
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, benchmark.
- score 100arxiv cs.CL (NLP)arxiv:2606.21123unread
A Multi-Agent Audit Framework for High-Stakes Reasoning: Evaluation and Interpretability in Clinical Mental Health Screening
Jingchen Ye, Yanpei Yu, Luyao Zhang · 2026-06-23
arXiv:2606. 21123v1 Announce Type: new Abstract: High-stakes reasoning tasks necessitate transparent and verifiable workflows, yet conventional single-model large language models (LLMs) often struggle with hallucination and low interpretability under zero-shot paradigms.
Read next because A Multi-Agent Audit Framework for High-Stakes Reasoning: Evaluation and Interpretability in Clinical Mental Health Screening 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, source, line, rate, chain, screen. Source: arxiv cs.CL (NLP).
arXiv:2606.21123v1 Announce Type: new Abstract: High-stakes reasoning tasks necessitate transparent and verifiable workflows, yet conventional single-model large language models (LLMs) often struggle with hallucination and low interpretability under zero-shot paradigms. To address this general AI challenge, we propose a Multi-Agent Audit Framework that simulates a collaborative, multi-step verification process. We empirically validate this architecture in the sensitive domain of clinical mental health screening using a modular LangChain workflow. Our framework decomposes the reasoning process into a Perception Agent, Knowledge Retrieval-Augmented Generation (RAG), Chain-of-Thought (CoT) clinical inference, and a critical Audit verification stage. We evaluated this framework on the DAIC-WOZ dataset using locally deployed open-source models. Experimental results demonstrate that our multi-agent pipeline significantly outperforms single-agent baselines, reducing the Mean Absolute Error (MAE) for PHQ-8 depression severity prediction from 5.35 to 5.02. By exposing cross-agent validation traces, the framework mitigates reasoning drift and provides highly interpretable diagnostic rationales, offering a generalizable paradigm for reliable AI-assisted decision support beyond isolated model scaling. We make data and code open access on GitHub for replicability.
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:2606.21098unread
LLM-Based Multi-Reference Evaluation for Efficient and Robust Assessment of Phrase Break Annotations
Younghan Park, Hoyeon Lee, Hawon Jeong, Jong-Hwan Kim · 2026-06-23
arXiv:2606. 21098v1 Announce Type: new Abstract: Reliable evaluation of phrase break annotations is crucial, as subtle variations in prosodic boundaries directly affect the clarity and naturalness of speech.
Read next because LLM-Based Multi-Reference Evaluation for Efficient and Robust Assessment of Phrase Break Annotations 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, phrase, rect, alignment, eval, rate, test, phrasings. Source: arxiv cs.CL (NLP).
arXiv:2606.21098v1 Announce Type: new Abstract: Reliable evaluation of phrase break annotations is crucial, as subtle variations in prosodic boundaries directly affect the clarity and naturalness of speech. However, existing approaches exhibit major limitations: single-reference evaluation assumes a unique gold phrasing for an utterance despite multiple valid phrasings, while human judgment, though flexible, is labor-intensive and unscalable. To address these, we propose LLM-based Multi-Reference Evaluation (LMRE) for phrase break annotations that models the one-to-many nature of prosodic phrasing and generates multiple valid phrasings from minimal demonstrations. On a Korean testbed of 1,356 annotations covering five strategies, LMRE shows stronger alignment with human judgment than single-reference evaluation in both acceptance behavior and score correlation. Our findings demonstrate that LMRE effectively achieves both scalability and multi-reference support, highlighting the potential of LLMs for evaluation in the speech domain.
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, evaluation.
- score 100arxiv cs.CL (NLP)arxiv:2606.21097unread
GRAG: Generic Response-Augmented Generation Framework for Personalized Conversational Systems
Junfeng Liu, Christopher T. Symons, Ranga Raju Vatsavai · 2026-06-23
arXiv:2606. 21097v1 Announce Type: new Abstract: Deploying highly capable personalized conversational agents in resource-constrained or privacy-sensitive environments remains a significant challenge.
Read next because GRAG: Generic Response-Augmented Generation Framework for Personalized Conversational Systems 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, source, line, rate, trained. Source: arxiv cs.CL (NLP).
arXiv:2606.21097v1 Announce Type: new Abstract: Deploying highly capable personalized conversational agents in resource-constrained or privacy-sensitive environments remains a significant challenge. We identify a fundamental bottleneck in the existing approaches: current training paradigms treat personalization and grounding as a single monolithic learning problem. Under these paradigms, language models are forced to simultaneously address what to say (content grounding) and how to say it in a user-specific way (personalization), which introduces significant computational and optimization challenges. Consequently, contextual grounding is often sacrificed for persona adherence, or vice versa, resulting in responses that are either weakly grounded in the conversational history or insufficiently personalized. In this work, we propose the Generic Response-Augmented Generation (GRAG) framework that decouples these competing objectives by leveraging offline, generic responses from high-capacity, general-purpose LLMs as a semantic and structural scaffold to guide the fine-tuning of smaller, task-specialized models seamlessly in resource-limited environments. By decoupling the content grounding from personalization, GRAG allows the model to focus exclusively on persona injection while remaining firmly anchored to the conversational context. We instantiate the GRAG in two post- and pre-fusion-based architectural variants and evaluate them on multiple benchmark conversational datasets that cover diverse personalization structures. Our results demonstrate that GRAG significantly outperforms state-of-the-art methods that do not use auxiliary scaffolding, yielding up to 47% improvements in ROUGE-2 and 36% in BLEU scores. Ultimately, GRAG offers a generalizable blueprint for building grounding-aware personalized conversational systems in resource-limited environments.
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:2606.21082unread
Scalable Hierarchical Attention Transformers for Multi-Turn Jailbreak Detection in Long Conversations
Chenhui Hu, Muhammed Salih, Sudipto Guha, Subramanian Srinivasan · 2026-06-23
arXiv:2606. 21082v1 Announce Type: new Abstract: Multi-turn jailbreaks can evade turn-level moderation by spreading unsafe intent across a dialogue through gradual escalation, reframing, and role manipulation.
Read next because Scalable Hierarchical Attention Transformers for Multi-Turn Jailbreak Detection in Long Conversations 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, eval, line, rate, full. Source: arxiv cs.CL (NLP).
arXiv:2606.21082v1 Announce Type: new Abstract: Multi-turn jailbreaks can evade turn-level moderation by spreading unsafe intent across a dialogue through gradual escalation, reframing, and role manipulation. We address multi-turn jailbreak detection as a conversation-level classification problem and introduce an efficient hierarchical detector that avoids expensive long-context concatenation while retaining cross-turn reasoning. The model encodes individual turns to form compact turn representations and applies a lightweight conversation module that captures dialogue dynamics and selectively attends to fine-grained evidence when needed. On a challenging evaluation benchmark of 14,038 conversations, our approach achieves an F1 of 0.9394, outperforming Claude Opus 4.7, the strongest competing baseline, by 0.07 while halving its false-positive rate. Ablation studies confirm that each architectural component contributes meaningfully, with combining cross-attention and self-attention in the conversation module yielding a 2.26 percentage point reduction in false-positive rate over the self-attention-only variant.
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:2606.21078unread
A Validation-Gated Mechanistic Account of Suicidality Detection in LLMs
Nafiz Ahmed, Sarah Sharif, Dingjing Shi, Mike Banad · 2026-06-23
arXiv:2606. 21078v1 Announce Type: new Abstract: Large language models are increasingly proposed for mental-health applications such as detecting suicidal content, raising the question of what they rely on.
Read next because A Validation-Gated Mechanistic Account of Suicidality Detection 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, word, latin, rect, line, rate, control, does. Source: arxiv cs.CL (NLP).
arXiv:2606.21078v1 Announce Type: new Abstract: Large language models are increasingly proposed for mental-health applications such as detecting suicidal content, raising the question of what they rely on. We study this mechanistically and use it to ask a narrower question: how to make a causal claim about a model's internal features more trustworthy. Our validation-gated framework, with suicidality detection as a case study, interprets a behavior only after the model is shown to perform it: a concept is admitted only once the model ranks it above a simple lexical baseline, and each subsequent property is tested against a matched control. This discipline yields negative as well as positive results. The gate rules out one task at the outset: on DeepSuiMind (Li et al. 2025), Llama-3.1-8B-Instruct cannot separate implicit suicidal intent from ordinary distress, so we do not analyze it. We turn to binary suicide detection, which it does perform. There we find a mid-network feature that appears semantic rather than keyword-based, is causally implicated in the decision (ablating it degrades the judgment; a random direction does not), is low-rank, and recurs across three model families and three suicide datasets. A register-matched control (suicide versus depression) suggests it tracks suicidality more specifically than general distress. Steering raises the model's response, but for unrelated questions too, so we treat it as necessary but not sufficient. The clearest pattern separates encoding from use: smaller models already represent suicidality, yet only larger ones appear to act on it. The positive evidence is English Reddit text, which limits the clinical reading.
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 cs.CL (NLP)arxiv:2606.21069unread
Quality and Agreement in Multilabel Emotion Annotation: A Case Study and Evaluation Framework
Emily \"Ohman, Anna Koufakou · 2026-06-23
arXiv:2606. 21069v1 Announce Type: new Abstract: Emotion annotation is inherently subjective, yet most NLP pipelines still assume "gold" labels, typically produced by majority voting, and treat annotator variation as noise.
Read next because Quality and Agreement in Multilabel Emotion Annotation: A Case Study and Evaluation Framework 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, soft, eval, line, model. Source: arxiv cs.CL (NLP).
arXiv:2606.21069v1 Announce Type: new Abstract: Emotion annotation is inherently subjective, yet most NLP pipelines still assume "gold" labels, typically produced by majority voting, and treat annotator variation as noise. In this paper, we present a multilabel emotion annotation case study and use it to examine how annotator behavior and aggregation choices affect both agreement estimates and downstream emotion classifiers. Rather than collapsing disagreement into a single label, we represent targets as soft vote-share labels (including an intensity-weighted variant) and evaluate models using both thresholded metrics (macro-/micro-F1) and probabilistic alignment (Bernoulli cross-entropy SoftBCE), alongside data-derived disagreement diagnostics. Across annotation regimes, we show that disagreement is structured and leaves measurable traces in model behavior: hard labels may maximize F1 metrics, while soft supervision yields predictions that better reflect empirical annotator variance and uncertainty. Our results provide practical guidance for designing, aggregating, and evaluating multilabel emotion datasets when multiple interpretations are plausible.
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 cs.CL (NLP)arxiv:2606.21066unread
Demographic Metadata as Construct-Irrelevant Noise in DistilBERT-Based Automated Essay Scoring
Teik Peng Ch'ng, Hui Na Chua · 2026-06-23
arXiv:2606. 21066v1 Announce Type: new Abstract: Automated Essay Scoring (AES) systems are increasingly used to support teachers in managing grading workloads and to provide a supplementary rater in large-scale assessments.
Read next because Demographic Metadata as Construct-Irrelevant Noise in DistilBERT-Based Automated Essay Scoring 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, token, line, rate, compare, trained. Source: arxiv cs.CL (NLP).
arXiv:2606.21066v1 Announce Type: new Abstract: Automated Essay Scoring (AES) systems are increasingly used to support teachers in managing grading workloads and to provide a supplementary rater in large-scale assessments. While human grading is frequently influenced by students' demographic characteristics, the efficacy of different strategies for integrating demographic metadata with textual input used to train AES models remains underexplored. This study investigates the impact of a specific multimodal fusion strategy - naive metadata concatenation - on the predictive accuracy, training convergence, and score parity of a DistilBERT-based AES model. A comparative analysis was conducted using the ASAP 2.0 dataset to evaluate a baseline model against an experimental model trained with input that concatenates tokenised text and demographic metadata using a naive multimodal fusion strategy. Evaluated via 10-fold cross-validation, the findings reveal that the early fusion of demographic metadata and the input significantly degrades the model's overall predictive accuracy. The baseline model achieved a Quadratic Weighted Kappa (QWK) of 0.727, which dropped to 0.656 upon integrating metadata. Furthermore, the experimental model exhibited higher validation loss (1.29) compared to the baseline model (1.25). The experimental model also displayed exacerbated scoring bias, reducing score parity instances from 15 to 12 out of 19 tests.
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:2606.21008unread
The Metanym Game: A Self-Contained, Self-Consistent LLM Peer-Community Benchmark for Structural Intelligence
David Nordfors · 2026-06-23
arXiv:2606. 21008v1 Announce Type: new Abstract: The metanym game is a competitive word game for LLMs that measures structural intelligence against established cognitive-science constructs.
Read next because The Metanym Game: A Self-Contained, Self-Consistent LLM Peer-Community Benchmark for Structural Intelligence 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, word, eval, rate, without, does, position, test. Source: arxiv cs.CL (NLP).
arXiv:2606.21008v1 Announce Type: new Abstract: The metanym game is a competitive word game for LLMs that measures structural intelligence against established cognitive-science constructs. No content is given in advance; the contestants create all of it -- a new kind of analogy test, analogical production falsifiable sentence by sentence, with no fixed test set to leak into training (contamination-resistant by construction). In the council-of-peers benchmark, the contestants also rate each other's creations. We introduce the first spectral solution, to our knowledge, to the wicked problem of benchmarking LLMs' factual accuracy without golden keys or oracle models: one singular value decomposition of the evaluators' ratings matrix yields their competence as both generators and judges of true statements at once. Competence on the subjective criteria comes from each judge's rating consistency as the yardstick shifts. The factual rating correlates with GPQA Diamond at Pearson r = 0.92. Scored separately, making and judging dissociate -- judging is the scarcer skill: the strongest generators are middling judges, the sharpest judge a mid-pack generator. To scale, the strongest players form a council that does the official benchmarking; its seats are contestable -- a stronger model earns one on the benchmark's own rating. The benchmark is entirely self-contained and self-consistent, a stable gauge over 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 benchmark.
- score 100arxiv cs.CL (NLP)arxiv:2606.20954unread
Learning What Not to Forget: Long-Horizon Agent Memory from a Few Kilobytes of Learning
Nusrat Jahan Lia, Aritra Mazumder · 2026-06-23
arXiv:2606. 20954v1 Announce Type: new Abstract: Long-running language-model systems accumulate interaction history that outgrows the context window, so they must continually evict.
Read next because Learning What Not to Forget: Long-Horizon Agent Memory from a Few Kilobytes of 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: code, text, under, eval, token, line, extraction, control. Source: arxiv cs.CL (NLP).
arXiv:2606.20954v1 Announce Type: new Abstract: Long-running language-model systems accumulate interaction history that outgrows the context window, so they must continually evict. When an eviction policy drops a load-bearing detail, for example an access token issued at login or a path the next call needs, the action fails. We present LRE (Learned Relevance Eviction), a few kilobytes, CPU-only, language-model-free scorer that learns which units of history are load-bearing and keeps them by verbatim extraction. Under a matched-budget comparison, in our experiment, no baseline dominates LRE on the accuracy-cost plane. On agents, LRE matches the accuracy of keeping the entire history overall. On the simplest tasks, it exceeds that no-eviction baseline by 27%, while requiring zero compressor calls and reducing peak context size by up to 52%. A controlled study trace shows LRE completes tasks where the others loop, finishing one such task in 37% fewer calls than keeping everything and solving 14 tasks where no other run policy does. On conversational memory, LRE outranks dense and token-pruning encoders at zero neural cost. In downstream evaluation, LRE gives the best budgeted answer quality on LoCoMo reading 68% fewer tokens. Its supervision can also be annotation-free: training only on the system's own behavior recovers 95% of the supervised scorer's effectiveness. We argue that, because memory eviction in LLM agents is a fidelity problem, it requires a deployable proactive policy where the future query is unavailable and exact state is decisive, and that cheap learned relevance can be sufficient.
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:2606.20936unread
Comparing Transformers and Hybrid Models at the Token Level
Yanhong Li, William Merrill · 2026-06-23
arXiv:2606. 20936v1 Announce Type: new Abstract: Hybrid language models that mix attention and recurrent layers have shown promise: theoretically, recurrent layers ameliorate the limitations of pure transformers on state tracking, and empirically, hybrids can outperform pure transformers in loss and downstream evaluations \citep{waleffe2024empirical,merrill2026olmohybrid}.
Read next because Comparing Transformers and Hybrid Models at the Token Level 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, word, class, under, eval, prefix, token, rate. Source: arxiv cs.CL (NLP).
arXiv:2606.20936v1 Announce Type: new Abstract: Hybrid language models that mix attention and recurrent layers have shown promise: theoretically, recurrent layers ameliorate the limitations of pure transformers on state tracking, and empirically, hybrids can outperform pure transformers in loss and downstream evaluations \citep{waleffe2024empirical,merrill2026olmohybrid}. Yet it remains unclear which data or capabilities drive these gains, and to what degree they reflect the theoretical advantages motivating hybrid models. We address this question using the open weights from Olmo 3 \citep{olmo2025olmo3} and Olmo Hybrid \citep{merrill2026olmohybrid}: we compare the loss of a matched transformer and hybrid at the same target tokens under the same prefixes, stratifying the results by natural token tags, copy features, delimiter structure, and controlled synthetic probes. The hybrid has lower loss on most tag families, but the gains are not uniform: they are largest for open-class content words and smaller for many closed-class function words. Across prose, code, and markup, the hybrid's loss advantage is larger on opening delimiters than on the corresponding closing delimiters, and nearly vanishes on repeated $n$-grams. Synthetic probes show the same split: the hybrid is favored on pronoun-memory and entity-tracking tasks, whereas the transformer is favored on bracket-matching tasks that require choosing closing delimiters. These patterns suggest that the recurrent layers in hybrids improve predictions that leverage the semantic state of a document, whereas attention helps on tokens predictable by $n$-gram copying or syntactic bracket matching. We conclude with proof-of-concept filtered evaluations showing how token-level decompositions can sharpen pretraining diagnostics for hybrid architectures.
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, evaluation.
- score 100arxiv cs.CL (NLP)arxiv:2606.20911unread
Latent Personal Memory: Represent personal memory as dynamic soft prompts
Debrup Das, Avinash Amballa, Yashas Malur Saidutta, Vijay Srinivasan, Vivek Kulkarni, Srinivas Chappidi · 2026-06-23
arXiv:2606. 20911v1 Announce Type: new Abstract: Personalizing large language models (LLMs) requires encoding long-term, user-specific behavioral patterns in a way that is computationally efficient, scalable, and compatible with a frozen base model.
Read next because Latent Personal Memory: Represent personal memory as dynamic soft 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: text, persona, soft, eval, rate, project, full, length. Source: arxiv cs.CL (NLP).
arXiv:2606.20911v1 Announce Type: new Abstract: Personalizing large language models (LLMs) requires encoding long-term, user-specific behavioral patterns in a way that is computationally efficient, scalable, and compatible with a frozen base model. We present Latent Personal Memory (LPM), a scalable framework that represents user-specific history as a compact, persistent matrix of N latent slots, that are interpretable. A shared cross-attention projection network maps these slots into dynamic, input-conditioned soft prompts that are prepended to the input of a frozen LLM. We evaluate LPM on PersonaMem v1 and LoCOMO benchmarks across Qwen3-1.7B, 4B, and 8B backbones. Results demonstrate that LPM outperforms LoRA and Prompt Tuning by up to 8.8% and 54.4% in overall accuracy respectively on PersonaMem v1, while reducing KV-cache usage by over 64x. On LoCoMo, LPM matches LoRA accuracy with 120x fewer trainable parameters. We also show that the efficiency of LPM grows with context length and outperforms full-context at 128K context length.
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:2606.20900unread
Storyline Trees: Hierarchical Representations for Long-Form Narratives
Litu Ou, Mirella Lapata · 2026-06-23
arXiv:2606. 20900v1 Announce Type: new Abstract: Long-form narratives are challenging for long-context models because their structure is implicit: events, characters, and plotlines interact across hundreds of pages without the explicit cues that guide navigation in structured documents.
Read next because Storyline Trees: Hierarchical Representations for Long-Form Narratives 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, without, trained, model. Source: arxiv cs.CL (NLP).
arXiv:2606.20900v1 Announce Type: new Abstract: Long-form narratives are challenging for long-context models because their structure is implicit: events, characters, and plotlines interact across hundreds of pages without the explicit cues that guide navigation in structured documents. We address this by constructing storyline trees, hierarchical representations that organize narratives from global themes and major plotlines to fine-grained events. We first segment chapters into contiguous narrative segments, or scenes, and use them as the basic units for tree construction. We then infer storyline trees through complementary top-down and bottom-up procedures that derive, refine, cluster, and summarize storylines at multiple levels of abstraction. We showcase the utility of this representation for question answering: storyline trees enable adaptive retrieval, allowing models to iteratively inspect high-level narrative structure and retrieve scene-level evidence on demand. Experiments on three long-context narrative QA benchmarks show that adaptive retrieval outperforms strong baselines, including post-trained long-context models and agentic chunk-based methods. Ablations confirm that scenes are more effective basic units than chapters or generic segmentation, and that gains persist under matched retrieval budgets
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:2606.20897unread
PeerCheck: Enhancing LLM-Generated Academic Reviews Towards Human-Level Quality
Zeyuan Chen, Ziqing Yang, Yihan Ma, Michael Backes, Yang Zhang · 2026-06-23
arXiv:2606. 20897v1 Announce Type: new Abstract: As academic submissions grow, the traditional peer review process struggles to keep up, raising concerns about quality and fairness.
Read next because PeerCheck: Enhancing LLM-Generated Academic Reviews Towards Human-Level Quality 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, rate, chain, language, model. Source: arxiv cs.CL (NLP).
arXiv:2606.20897v1 Announce Type: new Abstract: As academic submissions grow, the traditional peer review process struggles to keep up, raising concerns about quality and fairness. A trend of using large language models (LLMs) for assistance has emerged. In this work, we take a critical step toward improving the quality of LLM-generated reviews. We propose the PeerCheck framework, which investigates LLM-human review differences (RQ1) and explores methods to improve LLM-generated review quality (RQ2). We first analyzed the human-written reviews with reviews generated by various LLMs and found that LLMs and humans focus on different terms, e.g., LLMs prioritize theory while humans emphasize methodology and experiments. We further adopt prompt engineering, such as Chain-of-Thought (CoT), and utilize retrieval-augmented generation (RAG) to enhance the LLM-generated reviews towards human-level quality. We find CoT significantly improves the quality of LLM reviews, while we discover an unexpected "RAG paradox," i.e., experiments with RAG produce different results for various LLMs and, in some cases, even reduce review quality. Our comprehensive analysis of LLM-generated academic reviews illustrates both possibilities and limitations, contributing to a more effective, human-aligned review system. Our dataset is available on https://github.com/TrustAIRLab/PeerCheck.
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 limitation, limitations.
- score 100arxiv cs.CL (NLP)arxiv:2606.20770unread
Beyond 'One Language, One Script': Quantifying Orthographic Bias in Multilingual VLMs with PuMVR
Prabhjot Singh, Bhushan Pawar, Madhu Reddiboina · 2026-06-23
arXiv:2606. 20770v1 Announce Type: new Abstract: Current Vision-Language Models (VLMs) are celebrated for their multilingual capabilities, yet they operate under a flawed assumption: that one language corresponds to a single writing system.
Read next because Beyond 'One Language, One Script': Quantifying Orthographic Bias in Multilingual VLMs with PuMVR 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, alone, does, chain, capability, another. Source: arxiv cs.CL (NLP).
arXiv:2606.20770v1 Announce Type: new Abstract: Current Vision-Language Models (VLMs) are celebrated for their multilingual capabilities, yet they operate under a flawed assumption: that one language corresponds to a single writing system. This overlooks billions of users of multi-script languages like Punjabi, Serbian, Hindi-Urdu, Kurdish, among many others, for whom a model's capability may be fractured by orthographic bias. We introduce PuMVR (Punjabi Multimodal Visual Reasoning), the first benchmark designed to quantify script-dependent bias through 375 culturally grounded image-reasoning tasks across Punjabi's three active scripts (Gurmukhi, Shahmukhi, Roman). Evaluating 10 state-of-the-art VLMs, we expose a substantial Script Gap: models frequently solve visual puzzles in one script while failing identical tasks in another, with accuracy deltas reaching 16% and Script Consistency Rates (SCR) as low as 24.8%. Crucially, visual input boosts absolute performance but does not close this gap, the relative bias persists. Our analysis suggests reasoning patterns show limited cross-script transferability, and Chain-of-Thought pathways diverge based on script alone. We propose SCR as a core metric for script-agnostic evaluation, challenging current multilingual assessment paradigms and providing a framework for equitable AI.
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, evaluation, benchmark.
- score 100arxiv cs.CL (NLP)arxiv:2606.20740unread
VeriBound: PAC-Bayesian Generalization Bounds for Process Reward Models Trained with Formal Verification Tools
Amirul Rahman, Mohammed Sabih Alsharari · 2026-06-23
arXiv:2606. 20740v1 Announce Type: new Abstract: Process Reward Models (PRMs) provide step-level verification for Large Language Model (LLM) reasoning, yet their training data acquisition remains a bottleneck: human annotation is costly and Monte Carlo roll-out estimates are noisy.
Read next because VeriBound: PAC-Bayesian Generalization Bounds for Process Reward Models Trained with Formal Verification Tools 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, trained, test, language, model. Source: arxiv cs.CL (NLP).
arXiv:2606.20740v1 Announce Type: new Abstract: Process Reward Models (PRMs) provide step-level verification for Large Language Model (LLM) reasoning, yet their training data acquisition remains a bottleneck: human annotation is costly and Monte Carlo roll-out estimates are noisy. A recent approach, FOVER, trains PRMs on step-level error labels automatically annotated by formal verification tools such as Z3 and Isabelle, and empirically observes cross-task generalization from symbolic tasks to diverse reasoning benchmarks. However, this generalization phenomenon lacks any theoretical explanation, and no formal bounds exist on the generalization error, sample complexity, convergence rate, or downstream Best-of-K performance of such PRMs. We propose VeriBound, a theoretical framework that provides PAC-Bayesian generalization bounds for PRMs trained with formal verification tools. We establish four main results: (i) a PAC-Bayesian generalization bound that relates the empirical verification error on formal-verification-annotated training data to the expected error on unseen reasoning tasks, with the bound depending on the formal verification accuracy and the divergence between training and test task distributions; (ii) a sample complexity result showing that $O(d \log(d/\delta) / \epsilon^2)$ formal-verification-annotated examples suffice to achieve generalization error $\epsilon$ with probability $1-\delta$, where $d$ is the complexity of the PRM hypothesis class; (iii) a convergence analysis proving that PRM training with formal verification labels converges at a linear rate under $L$-smoothness and bounded variance conditions; and (iv) an error propagation bound that relates step-level verification error to Best-of-K performance degradation.
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.CL (NLP)arxiv:2606.20572unread
Investigating Linguistic Steering: An Analysis of Adjectival Effects Across Large Language Model Architectures
Lars Malmqvist · 2026-06-23
arXiv:2606. 20572v1 Announce Type: new Abstract: Achieving reliable control of Large Language Models (LLMs) requires a precise, scalable understanding of how they interpret linguistic cues.
Read next because Investigating Linguistic Steering: An Analysis of Adjectival Effects Across Large Language Model Architectures 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, alignment, line, rate, control, follow-up. Source: arxiv cs.CL (NLP).
arXiv:2606.20572v1 Announce Type: new Abstract: Achieving reliable control of Large Language Models (LLMs) requires a precise, scalable understanding of how they interpret linguistic cues. We introduce a rigorous framework using Shapley values to quantify the steering effect of individual adjectives on model performance, moving beyond anecdotal heuristics to principled attribution. Applying this method to 100 adjectives across a diverse suite of models (including o3, gpt-4o-mini, phi-3, llama-3-70b, and deepseek-r1) on the MMLU benchmark, we uncover several critical findings for AI alignment. First, we find that a small subset of adjectives act as disproportionately powerful "levers," yet their effects are not universal. Cross-model analysis reveals a "family effect": models of a shared lineage exhibit correlated sensitivity profiles, while architecturally distinct models react in a largely uncorrelated manner, challenging the notion of a one-size-fits-all prompting strategy. Second, focused follow-up studies demonstrate that the steering direction of these powerful adjectives is not intrinsic but is highly contingent on their syntactic role and position within the prompt. For larger models like gpt-4o-mini, we provide the first quantitative evidence of strong, non-additive interaction effects where adjectives can synergistically amplify, antagonistically dampen, or even reverse each other's impact. In contrast, smaller models like phi-3 exhibit a more literal and less compositional response. These results suggest that as models scale, their interpretation of prompts becomes more sophisticated but also less predictable, posing a significant challenge for robustly steering model behavior and highlighting the need for compositional and model-specific alignment techniques.
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:2606.21023unread
Demystifying Numerical Instability in LLM Inference: Achieving Reproducible Inference for Mission-Critical Tasks with HEAL
Zhenting Zhu, Lucas Thai, Shan Yu, Yicheng Liu, Yifan Qiao, Chenxi Wang, Harry Xu, Junyi Shu · 2026-06-23
arXiv:2606. 21023v1 Announce Type: new Abstract: As Large Language Models (LLMs) deploy into mission-critical domains (e.
Read next because Demystifying Numerical Instability in LLM Inference: Achieving Reproducible Inference for Mission-Critical Tasks with HEAL 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 "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: under, width, eval, line, rate, without, language, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.21023v1 Announce Type: new Abstract: As Large Language Models (LLMs) deploy into mission-critical domains (e.g., finance, medicine, and law), output reproducibility has become a strict system requirement. While practitioners use greedy decoding to eliminate algorithmic stochasticity, empirical deployments with 16-bit precisions still exhibit catastrophic output divergence across heterogeneous GPUs. Through SASS-level profiling, we reveal that this inconsistency is fundamentally driven by truncation errors introduced during downcasting at kernel boundaries. However, achieving reproducibility via a global FP32 pipeline incurs prohibitive system penalties: bypassing 16-bit hardware accelerators hurts compute efficiency, while upcasting the KV cache doubles memory overhead. To bridge this gap, we propose Hybrid Error ALleviation (HEAL), a targeted intervention that approximates FP32 precision while resolving hardware constraints through two targeted mechanisms. First, recognizing that floating-point formats underutilize their bit-width for Q, K, V tensors, HEAL applies INT16 quantization that preserves numerical stability without expanding the KV cache footprint. Second, HEAL synthesizes high-precision matrix multiplications via an algebraic error compensation strategy, executing entirely on high-throughput 16-bit Tensor Cores. To evaluate our approach practically, we introduce MCR-Bench, a benchmark targeting reproducibility in mission-critical tasks. HEAL achieves the same level of reproducibility on downstream tasks as the FP32 baseline while reducing the performance overhead by up to 7.1x.
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:2606.21022unread
Structure-Aware Graph Multi-Task Learning for Dynamic Sparse OD Demand Prediction
Ming Xu, Jiawei Cao · 2026-06-23
arXiv:2606. 21022v1 Announce Type: new Abstract: Origin-Destination (OD) demand prediction is fundamental to intelligent transportation systems, yet real-world OD flows are often dynamically sparse, long-tailed, and characterized by heterogeneous zero-flow patterns.
Read next because Structure-Aware Graph Multi-Task Learning for Dynamic Sparse OD Demand Prediction 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 "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation". Matching terms: line, rate, chen, compare, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.21022v1 Announce Type: new Abstract: Origin-Destination (OD) demand prediction is fundamental to intelligent transportation systems, yet real-world OD flows are often dynamically sparse, long-tailed, and characterized by heterogeneous zero-flow patterns. These properties make it difficult to distinguish whether an OD connection is active from how much demand it generates once activated. Many existing methods primarily treat OD prediction as a single flow regression task, which limits their ability to model low-frequency, intermittent, and long-tailed OD interactions. To address these challenges, we propose SAGMTL, a Structure-Aware Graph Multi-Task Learning framework for dynamic sparse OD demand prediction. SAGMTL decomposes OD prediction into structural state modeling and flow intensity estimation, jointly learning regional activity states, OD connection activity, and edge-level flow intensity within a unified framework. Specifically, a node-edge collaborative representation module captures regional semantics, temporal dynamics, and spatial priors through interactive node-edge updates, producing structure-aware representations for dynamic OD interactions. Based on these representations, SAGMTL estimates OD flows by jointly modeling stable demand patterns and short-term fluctuations. A multi-constraint objective further improves sparsity awareness and structural consistency. Experiments on three real-world urban mobility datasets from Beijing, Chengdu, and Nanjing show that SAGMTL achieves superior overall performance compared with state-of-the-art baselines. Further analysis demonstrates that explicitly modeling regional activity, connection states, and flow intensity improves the robustness of dynamic sparse OD demand prediction.
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 robustness.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20983unread
Physics-Guided Fully Convolutional Spatiotemporal Learning Toward Digital-Twin-Enabled Microstructure Evolution Prediction
Michael Trimboli, Wenxi Liu, Xianqi Li · 2026-06-23
arXiv:2606. 20983v1 Announce Type: new Abstract: Understanding and predicting microstructure evolution is central to materials design, yet purely data-driven spatiotemporal learning models often suffer from limited physical consistency and degraded long-term prediction accuracy.
Read next because Physics-Guided Fully Convolutional Spatiotemporal Learning Toward Digital-Twin-Enabled Microstructure Evolution Prediction 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, line, rate, compare, full, position, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20983v1 Announce Type: new Abstract: Understanding and predicting microstructure evolution is central to materials design, yet purely data-driven spatiotemporal learning models often suffer from limited physical consistency and degraded long-term prediction accuracy. In this work, we introduce a physics-guided fully convolutional spatiotemporal learning framework for microstructure evolution prediction. Unlike prior self-supervised approaches, the proposed method explicitly incorporates governing physical equations into the training objective, thereby encouraging the learned dynamics to remain consistent with known thermodynamic and kinetic laws. This physics-guided formulation improves predictive accuracy, long-horizon stability, and robustness across spatial resolutions and temporal prediction settings. Extensive experiments for spinodal decomposition demonstrate that incorporating physics-guided residual regularization leads to more faithful reproduction of microstructural morphology, statistics, and evolution trends compared with purely data-driven baselines. The proposed framework preserves the scalability and computational efficiency of fully convolutional architectures while bridging the gap between high-fidelity physics-based simulations and data-driven surrogate modeling, offering a reliable and efficient surrogate-modeling step toward digital-twin-enabled microstructure evolution prediction.
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.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20976unread
Detecting Satellites in Radio-Frequency Data via Semi-Supervised Learning
Cade W. Trotter, Maksim E. Eren, Justin C. Holmes, J. Brent Parham, David Ewing, Boian S. Alexandrov, Gian Luca Delzanno · 2026-06-23
arXiv:2606. 20976v1 Announce Type: new Abstract: Radio-frequency (RF) monitoring is essential for space domain awareness, but it often generates large, variable, and sparsely populated datasets with few labels.
Read next because Detecting Satellites in Radio-Frequency Data via Semi-Supervised Learning 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, rate, factor, test, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20976v1 Announce Type: new Abstract: Radio-frequency (RF) monitoring is essential for space domain awareness, but it often generates large, variable, and sparsely populated datasets with few labels. These observations can capture satellites, space debris, and the ionospheric background, yet interpreting them typically requires specialized subject-matter expertise. Supervised deep learning methods can perform well on labeled RF data, but they require many annotated examples and may need careful retraining as RF conditions change. Semi-supervised approaches offer a practical alternative for limited-data settings by using unlabeled observations to reveal latent patterns that experts can interpret. In this paper, we present a semi-supervised RF detection and classification workflow for satellite monitoring that combines Non-negative Matrix Factorization with automatic model determination (NMFk), expert-guided cluster interpretation, and classifier-based prediction. We first represent RF observations as a non-negative feature matrix and apply NMFk to estimate the number of clusters that best captures patterns in the unlabeled data. Subject-matter experts then assign physical meaning to the resulting clusters, including satellite detections, ionospheric environmental conditions, and other RF event categories. Finally, we train a classifier on these interpreted clusters to evaluate performance on a test set and categorize future observations. This pipeline reduces reliance on large pre-labeled datasets by pairing unsupervised factorization with expert interpretation, enabling an interpretable and transferable methodology for detecting, observing, and classifying behavior in RF data.
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 negative.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20961unread
Is Our Benchmark Enough? An Analysis of Continual Learning for MLLMs
Van-Tuan Tran, Shruthi Gowda, Merim Dzaferagic, Marco Ruffini · 2026-06-23
arXiv:2606. 20961v1 Announce Type: new Abstract: Continual adaptation is essential for multimodal large language models (MLLMs) deployed across evolving domains, but the state-of-the-art MR-LoRA method highly relies on the assumption that a MLLM-based router is necessary to process complex multimodal inputs.
Read next because Is Our Benchmark Enough? An Analysis of Continual Learning for MLLMs 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, does, trained, lora, language, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20961v1 Announce Type: new Abstract: Continual adaptation is essential for multimodal large language models (MLLMs) deployed across evolving domains, but the state-of-the-art MR-LoRA method highly relies on the assumption that a MLLM-based router is necessary to process complex multimodal inputs. This paper revisits this claim on the MLLM-CL benchmark and argues for two claims. \textbf{First}, routing does not require an MLLM: a simple training-free, replay-free ptotypical routing method (\textsc{RePRo}), uses frozen pretrained features and task prototypes to match the MLLM-based router of MR-LoRA at far lower computational cost. \textbf{Second}, shared experts do not improve continual learning for MLLMs, despite their theoretical appeal. We show that these findings arise from two structural limitations of MLLM-CL: (1) its tasks are \textbf{highly separable} in representation space, and (2) its fixed task order makes conclusions \textbf{sensitive to a single curriculum} rather than robust across diverse continual-learning trajectories. As a result, the benchmark primarily rewards learning in isolation rather than genuine continual transfer. This motivates a new design for future benchmarks of continual MLLM learning, with overlapping task manifolds, multiple task orders, fine-grained domain shifts, and evaluation protocols that reward forward transfer as well as retention.
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, evaluation, benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20959unread
Right Knowledge, Wrong Answer: Test-Time Steering for Temporal Fact Conflicts in Open-Weight Language Models
Elias Hossain, Sourav Saha, Umesh Chandra Biswas, Sanjeda Sara Jennifer · 2026-06-23
arXiv:2606. 20959v1 Announce Type: new Abstract: Large language models can store both outdated facts and newer superseding facts in their parameters, but standard prompting may still elicit the outdated answer.
Read next because Right Knowledge, Wrong Answer: Test-Time Steering for Temporal Fact Conflicts in Open-Weight Language Models 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: wrong, eval, line, rate, without, stage, test, language. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20959v1 Announce Type: new Abstract: Large language models can store both outdated facts and newer superseding facts in their parameters, but standard prompting may still elicit the outdated answer. We formalize this problem as Parametric Temporal Conflict (PTC) and introduce Temporal Attractor Steering (TAS), a three-stage test-time intervention that detects likely conflicts, identifies a conflict-critical layer, and steers hidden states toward newer-fact representations without retraining or external retrieval. We construct an 8,746-record verified benchmark across five Wikidata relations and evaluate four open-weight language models from three families: Qwen-2.5-1.5B/7B, Mistral-7B-v0.3, and Llama-3.1-8B. Single-layer activation patching achieves answer-flip rates of 0.72-0.85 across all models. End-to-end TAS resolves 29-57% of PTC cases while preserving 85-99% accuracy on non-conflict queries, outperforming a matched ITI baseline on three of four models. These results show that outdated parametric knowledge can be selectively overridden at inference time.
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:2606.20933unread
Towards Robust Training in NNGPT AutoML Pipeline: A Loss-Optimizer Pairing Selection Study
Anton Abramochkin, Radu Timofte, Dmitry Ignatov · 2026-06-23
arXiv:2606. 20933v1 Announce Type: new Abstract: The choice of loss function and optimizer is an important decision, that shapes further model training.
Read next because Towards Robust Training in NNGPT AutoML Pipeline: A Loss-Optimizer Pairing Selection Study 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, eval, source, line, rate, recipe, full. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20933v1 Announce Type: new Abstract: The choice of loss function and optimizer is an important decision, that shapes further model training. Yet automated architecture search pipelines (AutoML) benefits significantly more from the optimal pairing selection and vice versa. This paper investigates whether a single recipe is sufficient for heterogeneous architecture pools, or whether the optimal pairing varies across structurally diverse models. We conduct a systematic empirical study of all $3 \times 6 = 18$ combinations of six optimizers (SGD+Momentum, Adam, AdamW, RMSprop, Adagrad, Adadelta), paired with three loss functions: Cross-Entropy (CEL), Negative Log-Likelihood (NLL), and the recently introduced genetically evolved NGL loss across the base models presented in LEMUR heterogeneous architecture pool on six image classification datasets (CelebA-Gender, CIFAR-10, CIFAR-100, ImageNette, MNIST, SVHN). The 18 loss-optimizer configurations are applied to each of the 33 compatible base architectures taken from the LEMUR pool, resulting in 594 variants that were generated fully automatically by a source-level injection pipeline and evaluated under fixed hyperparameters, ensuring that observed accuracy differences are attributable solely to the loss-optimizer pairing. Our results confirm that no single pairing is universally optimal. Cross-Entropy with Adam or AdamW is the most robust choice across architecture families and datasets. NGL is a competitive alternative to CEL on standard convolutional classifiers, but only when paired with adaptive optimizers; it degrades substantially with SGD or accumulation-based methods. Adagrad and Adadelta consistently underperform under fixed hyperparameters regardless of loss function, highlighting their sensitivity to learning rate tuning. These findings provide actionable guidance for loss-optimizer selection within NNGPT Framework.
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 negative.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20920unread
$\Omega$: Operator-based Mixture Ensemble for Generative Assimilation
Pouria Behnoudfar, Nan Chen · 2026-06-23
arXiv:2606. 20920v1 Announce Type: new Abstract: Characterizing non-Gaussian posterior distributions in partially observed high-dimensional nonlinear systems remains a fundamental challenge in data assimilation.
Read next because $\Omega$: Operator-based Mixture Ensemble for Generative Assimilation 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, alone, full, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20920v1 Announce Type: new Abstract: Characterizing non-Gaussian posterior distributions in partially observed high-dimensional nonlinear systems remains a fundamental challenge in data assimilation. Ensemble Kalman filters rely on Gaussian approximations that can be inaccurate for strongly non-Gaussian posteriors, whereas particle filters suffer from severe scalability limitations. Recent score-based generative approaches improve posterior characterization but typically require supervised training with ground-truth posterior samples, which are unavailable in most practical applications. We introduce $\Omega$ (Operator-based Mixture Ensemble for Generative Assimilation), a scalable framework that integrates conditional Gaussian surrogate modeling, unsupervised score learning, and generative sampling. The conditional Gaussian surrogate provides a nonlinear non-Gaussian baseline approximation while admitting closed-form conditional posterior distributions for the unresolved variables. First, $\Omega$ exploits these closed-form conditional distributions to analytically recover the high-dimensional unobserved component, reducing computational cost and mitigating the curse of dimensionality. Second, $\Omega$ learns only the residual discrepancy beyond an analytical baseline through denoising score matching using ensemble trajectories alone, eliminating the need for ground-truth posterior samples and substantially reducing the learning burden. Third, $\Omega$ reconstructs the full non-Gaussian posterior distribution of both observed and unobserved variables via a Gaussian mixture representation, capturing multimodal, skewed, and heavy-tailed statistics. Finally, $\Omega$ employs annealed Langevin sampling to iteratively refine ensemble members from the baseline toward the target posterior. $\Omega$ is validated on several turbulent models with intermittency and extreme events, consistently improving posterior accuracy.
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.LG (Machine Learning)arxiv:2606.20916unread
Physics-Guided Dual-Stream Heterogeneous Graph Neural Network for Predicting Full-Field Structural Response of Stiffened Panels
Yuecheng Cai, Jasmin Jelovica · 2026-06-23
arXiv:2606. 20916v1 Announce Type: new Abstract: Iterative design and optimization of large, complex structures require fast and accurate prediction of stress, displacement, and other fields.
Read next because Physics-Guided Dual-Stream Heterogeneous Graph Neural Network for Predicting Full-Field Structural Response of Stiffened Panels 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, line, rate, compare, full, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20916v1 Announce Type: new Abstract: Iterative design and optimization of large, complex structures require fast and accurate prediction of stress, displacement, and other fields. Finite element analysis (FEA) is computationally expensive for this task. Existing neural network surrogates often struggle with varying topologies and complex boundary conditions. This study proposes the novel Dual-Stream Heterogeneous Graph Neural Network (DS-HGNN) for full-field stress and displacement prediction in thin-walled structures, demonstrated on box beams made of stiffened panels. DS-HGNN operates on panel-level heterogeneous graph representations and introduces physics-guided edge states initialized from edge types, spatial information, and boundary kinematics. These states are updated through dual-stream message passing that separates longitudinal and transverse structural information while allowing cross-stream exchange. Geometry and loading effects are incorporated through Feature-wise Linear Modulation (FiLM)-conditioned 1-D spectral convolutions, and physical fields are reconstructed using a spectral-bypass low-rank readout. The model is evaluated on stiffened panel datasets with different geometries, boundary kinematics, loading conditions, and material nonlinear responses. DS-HGNN achieves the lowest stress and displacement RMSE compared with six benchmark heterogeneous graph neural network models. It also reaches comparable accuracy to the strongest benchmark models using 19%-38% fewer training samples. A targeted evaluation further shows that DS-HGNN captures yield and post-yield stress features.
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.LG (Machine Learning)arxiv:2606.20906unread
MMGNN: Multi-level, multi-color graph neural networks for molecular property prediction
Trung Nguyen, Duc Duy Nguyen · 2026-06-23
arXiv:2606. 20906v1 Announce Type: new Abstract: Molecular message-passing neural networks commonly propagate chemically diverse interactions through a single graph, which may mix interaction-specific signals and require deep propagation to capture long-range effects.
Read next because MMGNN: Multi-level, multi-color graph neural networks for molecular property prediction 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, propagate, position. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20906v1 Announce Type: new Abstract: Molecular message-passing neural networks commonly propagate chemically diverse interactions through a single graph, which may mix interaction-specific signals and require deep propagation to capture long-range effects. We introduce the Multi-level, Multi-color Graph Neural Network (MMGNN), a hierarchical framework that decomposes a molecular graph into overlapping atom-type-pair-specific subgraphs while preserving atom-level resolution. MMGNN-2D constructs chemical-colored subgraphs from covalent connectivity, whereas MMGNN-3D constructs geometric-colored subgraphs from spatial proximity and augments their edges with distance, angular, and torsional descriptors. Both variants apply a shared communicative message-passing backbone to each subgraph and combine the resulting representations through atom-wise aggregation and molecular readout. We evaluated MMGNN on five classification and three regression benchmarks from MoleculeNet using common scaffold splits and five independent runs. MMGNN-2D achieved the highest macro-average AUC-ROC of 0.838 across the classification datasets and the lowest RMSE on ESOL (0.803). MMGNN-3D obtained the highest mean AUC-ROC on BBBP (0.956) and the lowest RMSE on FreeSolv (1.793), indicating complementary strengths of topological and geometric representations. Structural and leave-one-out analyses further illustrate how the subgraph decomposition affects learned representations and atom-type-pair sensitivities. These results support overlapping interaction-specific graph decomposition as a competitive strategy for molecular property prediction.
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.LG (Machine Learning)arxiv:2606.20889unread
Temporal Causal Prior-Data Fitted Networks for Panel Data with Learned Reliability Signals
Shravan Talupula, Saurabh Sharma · 2026-06-23
arXiv:2606. 20889v1 Announce Type: new Abstract: Estimating causal effects in industrial time series requires handling temporal dynamics, time-varying treatments, and unobserved confounders.
Read next because Temporal Causal Prior-Data Fitted Networks for Panel Data with Learned Reliability Signals 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, latin, rect, under, correct, token, rate, control. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20889v1 Announce Type: new Abstract: Estimating causal effects in industrial time series requires handling temporal dynamics, time-varying treatments, and unobserved confounders. Existing causal foundation models (CausalPFN, CausalFM) operate only on static cross-sectional data; neural temporal methods (CRN, G-Net) require per-dataset training; and concurrent temporal-PFN proposals have not been demonstrated at industrial scale. None output explicit per-pair reliability signals alongside their CATE estimates. We introduce Temporal Causal Prior-Data Fitted Networks (TCPFN), a foundation model for zero-shot temporal causal discovery with learned reliability signals. TCPFN makes four contributions: (1) a Causal Judgment Head that jointly predicts null-effect probability, confounding strength, identifiability, mediation fraction, and causal regime; (2) a mixed training prior covering six causal regimes (independent, direct, confounded, mediated, time-varying confounded, feedback) plus CausalFM-style front-door and instrumental-variable priors; (3) a discrete-token panel-data architecture with cross-attention masking that prevents inter-horizon leakage; (4) zero-shot inference at industrial scale via FAISS-based context selection and one-step posterior correction. On 19 benchmark datasets across five domains, TCPFN achieves competitive zero-shot causal discovery: AUROC 0.96 on Tennessee Eastman, 0.93 on SWaT, 0.98 on Causal Rivers, 0.97 on CAUSRCA. The null detector reaches NullF1 0.94, AUROC 0.99. TCPFN scales to V=1,275 on a proprietary Kraft pulp-and-paper dataset in 6 hours on a single GPU; PCMCI, a CPU-only library, on a V=666 sub-panel of the same data took 81.5 hours, extrapolating by O(V^2) to ~12.5 days at V=1,275. TCPFN's top edges identify cross-subsystem causal relationships while PCMCI's surface within-instrument controller-measurement coupling -- a scalability case study.
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, benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20874unread
Machine Learning Classification of Cryopathy Syndromes: A Comprehensive Comparative Study
Nataliya Shakhovska, Valentyna Chopyak, Ivan Izonin, Vira Haievska · 2026-06-23
arXiv:2606. 20874v1 Announce Type: new Abstract: Cryopathy syndromes are difficult to classify because laboratory patterns often overlap across diagnostic categories, while some diagnoses are rare.
Read next because Machine Learning Classification of Cryopathy Syndromes: A Comprehensive Comparative Study 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, soft, eval, rate, compare, test, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20874v1 Announce Type: new Abstract: Cryopathy syndromes are difficult to classify because laboratory patterns often overlap across diagnostic categories, while some diagnoses are rare. This makes routine interpretation of cryoglobulin-related tests challenging and increases dependence on expert judgment. The aim of this study was to develop and compare machine learning approaches for automated classification of cryopathy syndromes from laboratory data and to identify a practical strategy for clinical decision support. Methods: We analysed laboratory records from 2,686 patients assigned to 14 diagnostic categories. The dataset included demographic variables, cryoglobulin measurements, precipitation tests, and hemagglutinin and hemolysin titers. Data preprocessing included cleaning, encoding, imputation, normalization, and construction of clinically informed interaction features. We evaluated 12 modelling strategies, including Random Forest, Gradient Boosted Trees, Multi-Layer Perceptron, soft-voting ensembles, class balancing with Synthetic Minority Over-sampling Technique, hierarchical classification, period-aware models, targeted binary classifiers, and probability calibration. Performance was assessed using stratified train-test evaluation and stratified 5-fold cross-validation. The main metrics were macro-averaged F1 score, accuracy, Top-3 accuracy, and expected calibration error. The overall task proved difficult because of marked class imbalance and clinical overlap between diagnoses. The best multiclass performance was achieved by a soft-voting ensemble of Random Forest and Gradient Boosted Trees. Cross-validation confirmed stable performance for the balanced Random Forest model. Tree-based methods consistently outperformed the neural network model. Feature engineering improved discrimination, and the most informative predictors were derived cryoglobulin-based interaction features.
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 cs.LG (Machine Learning)arxiv:2606.20820unread
CELEUS: Certifiable and Efficient LLM Evaluation via E-Processes
Zhijian Zhou, Zesheng Ye, Zhaorun Chen, Bo Li, Feng Liu · 2026-06-23
arXiv:2606. 20820v1 Announce Type: new Abstract: Can we trust evaluation scores to capture an LLM's true real-world performance?
Read next because CELEUS: Certifiable and Efficient LLM Evaluation via E-Processes 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 "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: width, eval, line, rate, factor. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20820v1 Announce Type: new Abstract: Can we trust evaluation scores to capture an LLM's true real-world performance? Certifiable evaluation answers this question by providing guarantee for LLM evaluation. In particular, existing methods sequentially curate evaluation samples and keep updating confidence intervals (CIs) that cover the true performance with high probability (e.g., 95%) until some conditions are satisfied, e.g., the CI width reaches a target precision. However, existing methods are not generally anytime-valid: the claimed coverage (e.g., 95%) may fail when CIs are repeatedly updated and used to decide when to stop, leaving a gap between theoretical rigor and practice. This paper bridges this gap by proposing Celeus, a Certifiable framework for Efficient LLM evaluation, which leverages E-processes to build anytime-valid CIs. Concretely, we propose signals that combine two ingredients: (i) Uncertainty-guided sampling to select informative samples for evaluation, and (ii) Surrogate-assisted approximations for unevaluated samples. We prove that such signals remain unbiased for the evaluation score conditional on the past, enabling statistically-grounded and anytime-valid $e$-process CIs. More importantly, the two ingredients reduce estimation variance and help reach the target precision with fewer evaluated samples. We also prove that CIs obtained by Celeus can shrink at a near-parametric rate up to logarithmic factors and analyze the oracle variance-optimal sampling rule that motivates the empirical uncertainty-guided one. Experiments show that Celeus reaches the target precision using 54-62% fewer evaluated samples than baselines, while preserving anytime-valid coverage.
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, evaluation.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20812unread
B[FM]$^2$: Brain Foundation Model via Flow Matching with SplitUNet
Jaedong Hwang, Kathleen Zhang, Wei Dai, Konstantinos Kontras, Maarten Vanmarcke, Maarten De Vos, Ila Fiete, Paul Pu Liang · 2026-06-23
arXiv:2606. 20812v1 Announce Type: new Abstract: EEG foundation models can learn generalizable representations from large-scale EEG corpora to enable single-backbone transfer across diverse clinical and brain-computer interface tasks.
Read next because B[FM]$^2$: Brain Foundation Model via Flow Matching with SplitUNet 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, rect, token, rate, project, without, factor. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20812v1 Announce Type: new Abstract: EEG foundation models can learn generalizable representations from large-scale EEG corpora to enable single-backbone transfer across diverse clinical and brain-computer interface tasks. Existing models typically discretize the continuous multi-channel EEG waveform into patches or codebook tokens and train a transformer with masked self-supervision. Recognizing that this discretization fragments continuous brain rhythms and obscures fine-grained temporal dynamics, we present B[FM]$^2$(Brain Foundation Model via Flow Matching), whose inductive bias aligns with the data by pretraining directly on the raw signal using continuous-time flow matching without patches, tokenization, or masking. However, multi-channel EEG signals pose an architectural challenge for flow matching: time is densely sampled and highly autocorrelated (thousands of timepoints), while the electrode axis is short (tens of channels) at distinct scalp positions. To address this time-electrode asymmetry, we introduce SplitUNet, a velocity network that factorizes each block into separate 1D temporal and 1D electrode convolutions and downsamples only along time, preserving electrode topology throughout the hierarchy. B[FM]$^2$ sets a new state of the art on 7 of 9 standard downstream EEG classification tasks, using a pretraining budget of only 36,895 segments ($\approx$ 307h), 1-2 orders of magnitude ($\approx$ 30x) less than required by existing EEG foundation models. Further, it generates synthetic EEGs that two board-certified neurologists cannot distinguish from brain data (Cohen's $\kappa =$ -0.096). https://jd730.github.io/projects/BFM2
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.LG (Machine Learning)arxiv:2606.20771unread
ELADO: Elliptic PDE Assessment Datasets for Operator Learning
Frank Ehebrecht, Toni Scharle, Martin Atzmueller · 2026-06-23
arXiv:2606. 20771v1 Announce Type: new Abstract: We introduce ELADO (Elliptic PDE Assessment Datasets for Operator Learning), a systematic benchmark suite constructed to show and quantify failure modes of neural operator architectures when learning solution operators of elliptic PDEs.
Read next because ELADO: Elliptic PDE Assessment Datasets for Operator 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 "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, control. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20771v1 Announce Type: new Abstract: We introduce ELADO (Elliptic PDE Assessment Datasets for Operator Learning), a systematic benchmark suite constructed to show and quantify failure modes of neural operator architectures when learning solution operators of elliptic PDEs. While the benchmarks of existing datasets focus on average case performance, the ELADO datasets are constructed to highlight challenges that arise naturally in elliptic PDE problems. In particular, we construct several datasets built around Poisson's equation and the Helmholtz equation, each with non-constant coefficients. We define a controllable data-generating process to create datasets, that are designed to isolate a distinct source of difficulty. Specifically, these are (1) heavy-tailed solution distributions arising from light-tailed coefficient field distributions, (2) spectral distribution shift of the input data, (3) heavy-tailed distributions in the frequency domain of solutions, arising from light-tailed coefficient field distributions, (4) input sensitivity of learned operators, quantified by an empirical local Lipschitz analysis, and (5) the effect of input signal complexity on prediction accuracy under controlled amplitude normalization. We evaluate several neural operator architectures across all datasets and show that heavy-tailed targets, spectral shift, and input sensitivity each cause substantial degradation of the prediction accuracy that standard datasets and metrics (e.g., the mean relative $L^2$ error) may obscure.
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, benchmark.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20757unread
Evidential Fusion Network for Multimodal Survival Prediction under Missing Modalities
Yucheng Xing, Hailan Mo, Zi Wang, Ling Huang, Mengling Feng · 2026-06-23
arXiv:2606. 20757v1 Announce Type: new Abstract: Recent multimodal survival prediction models have demonstrated strong predictive performance by leveraging complementary information across modalities.
Read next because Evidential Fusion Network for Multimodal Survival Prediction under Missing Modalities 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, under, rate, without, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20757v1 Announce Type: new Abstract: Recent multimodal survival prediction models have demonstrated strong predictive performance by leveraging complementary information across modalities. However, such models generally assume data completeness and exhibit limited robustness toward missing modalities, which are frequently encountered in real-world clinical settings. We propose the Evidential Missing Modality Survival Fusion (EMMS) model for multimodal survival prediction under missing modalities. EMMS offers a straightforward, computationally effective approach to survival analysis without requiring a generative phase for missing data. By employing Dempster-Shafer theory and Gaussian Random Fuzzy Numbers for multimodal decision fusion, it considers both aleatoric and epistemic uncertainty alongside modality reliability for fusion. Moreover, the model treats missing modalities as vacuous evidence, preventing interference with available inputs and naturally reflecting increased uncertainty and calibrated predictions. Extensive experiments on four cancer datasets demonstrate state-of-the-art performance while providing calibrated and interpretable uncertainty estimates under incomplete multimodal observations, without introducing additional computational overhead.
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.LG (Machine Learning)arxiv:2606.20673unread
NeuroShield: A Device-Agnostic Foundation Model for EEG Authentication
Matin Fallahi, Patricia Arias-Cabarcos, Thorsten Strufe · 2026-06-23
arXiv:2606. 20673v1 Announce Type: new Abstract: A central challenge in EEG authentication is that models are typically tied to the acquisition settings in which they are trained.
Read next because NeuroShield: A Device-Agnostic Foundation Model for EEG Authentication 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, source, rate, trained, length, stage, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20673v1 Announce Type: new Abstract: A central challenge in EEG authentication is that models are typically tied to the acquisition settings in which they are trained. In particular, variations in headset hardware, channel layout, and signal duration create heterogeneous recordings that existing models are not designed to handle, causing each new headset or dataset to be treated as a separate model-development problem. This fragmentation limits multi-dataset learning, hinders knowledge transfer, and reduces model reusability. To address this limitation, we present NeuroShield, a reusable foundation model for EEG authentication that learns identity-discriminative embeddings from variable-channel and variable-length EEG recordings through a dual-stage transformer architecture. We pretrain NeuroShield on three public EEG datasets comprising 15{,}762 subjects and 28{,}116 sessions, and evaluate transfer on two unseen downstream datasets. Our evaluations show that, after fine-tuning, NeuroShield reduces equal error rate by 0.44--8.06 percentage points relative to the state of the art. NeuroShield further generalizes to segments longer than those seen during training and operates across channel layouts not encountered during pretraining. These results establish NeuroShield as a reusable and adaptable EEG identity encoder across heterogeneous recording settings. We release NeuroShield as open source to support reproducibility and community adoption.
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, evaluation.
- score 100arxiv cs.LG (Machine Learning)arxiv:2606.20670unread
Towards CSI-Native Foundation Models: A Channel-Adaptive Roadmap for 6G
Chenyu Zhang, Xinchen Lyu, Chenshan Ren, Shuhan Liu, Qimei Cui · 2026-06-23
arXiv:2606. 20670v1 Announce Type: new Abstract: Wireless foundation models offer a path toward reusable channel state information (CSI) intelligence for sixth-generation (6G) systems.
Read next because Towards CSI-Native Foundation Models: A Channel-Adaptive Roadmap for 6G 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 "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: under, eval, token, rate, control, position, model. Source: arxiv cs.LG (Machine Learning).
arXiv:2606.20670v1 Announce Type: new Abstract: Wireless foundation models offer a path toward reusable channel state information (CSI) intelligence for sixth-generation (6G) systems. However, existing generic-backbone adaptation and CSI pretraining methods often treat CSI as task tensors rather than propagation-conditioned channel responses, thereby failing to capture the intrinsic time-frequency-spatial geometry of wireless environments. This paper presents a channel-adaptive roadmap toward CSI-native foundation models, proposing a unified framework that aligns pretraining, positional modeling, and attention control with three channel requirements: scale-aware heterogeneous exposure, physical time-frequency-antenna coordinates, and correlation-bounded token interaction. Extensive experiments demonstrate the superiority of the proposed framework across three dimensions: zero-shot generalization, reducing NMSE by more than 4 dB across spatial-temporal-frequency tasks; scale extrapolation, yielding up to a 5.4 dB gain under 8 times unseen antenna scaling; and inference efficiency, accelerating mobility-aware processing by up to 18.8%. A system-level evaluation with Sionna SYS further shows that the proposed framework uses only 7.01% of dense-pilot overhead, reaches -18.64 dB average NMSE, and improves average net spectral efficiency by 36.6% over dense LMMSE and 15.5% over WiFo, indicating that CSI-native representation learning can support pilot-efficient radio access.
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 stat.ML (Machine Learning)arxiv:2606.20814unread
What Shapes Emergent Misalignment? Insights from Training Dynamics, Model Priors, and Data
Yuchen Zhang, Anietta Weckauff, Diego Garcia-Olano, Maksym Andriushchenko · 2026-06-23
arXiv:2606. 20814v1 Announce Type: cross Abstract: Emergent misalignment (EM) is a phenomenon in which models generalize with narrow fine-tuning, leading to broad (yet uneven) misalignment across evaluation questions.
Read next because What Shapes Emergent Misalignment? Insights from Training Dynamics, Model Priors, and 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: strong, rect, alignment, eval, token, rate, compare, control. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.20814v1 Announce Type: cross Abstract: Emergent misalignment (EM) is a phenomenon in which models generalize with narrow fine-tuning, leading to broad (yet uneven) misalignment across evaluation questions. We study EM and its variability directly through the components of fine-tuning: training dynamics, model priors, and data. (1) We first explored how in-domain training loss relates to out-of-domain alignment scores across datasets and model families. Then, we tried to induce potential alternative local minima through different learning schedules for one narrow fine-tuning, but did not find strong runs with better broad alignment scores conditioned on similar or lower training loss. (2) We found that although the mean and standard deviations of the misaligned model scores are usually statistically different from those of the pre-trained model, there are some potential signals on overall positive correlation. The evaluation prompt-only activations from both the pre-trained and the original instruct models (prior to narrow fine-tuning) could predict fine-grained alignment scores after narrow fine-tuning. (3) Finally, we compared activation deltas before and after narrow fine-tuning and found moderate-to-high subspace overlap and similarity between the resulting activation shifts for training and evaluation prompts. Subspace overlaps between training and evaluation prompt activations correlate with their shifts' similarities when measuring with the last prompt-token activations. The train-evaluation data prompt overlap is controlled against overlap computed from random vectors and evaluation prompts activations.
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:2606.20624unread
In LLM Reasoning, there is Irrationality on top of Value Misalignment
Kejiang Qian, Fengxiang He · 2026-06-23
arXiv:2606. 20624v1 Announce Type: new Abstract: Significant progress has been made in aligning LLMs with target value functions.
Read next because In LLM Reasoning, there is Irrationality on top of Value Misalignment 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, alignment, eval, rate, candidates, candidate, model. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20624v1 Announce Type: new Abstract: Significant progress has been made in aligning LLMs with target value functions. We argue that, even when an LLM has been well aligned in (post-)training, it may still fail to maximise the aligned value in reasoning. We mathematically formalise this gap as rational value risk: the utility discrepancy between a model's deployed reasoning strategy and its rational counterpart, which is defined to be the responses that maximise expected utility in the steepest direction. The estimation error of rational value risk is further decomposed into three components from finite candidates, finite prompts, and imperfect verifiers. Extensive experiments are conducted, covering models Llama-3.1, Qwen-2.5, T{\"}ulu-3 families (7B-72B), GPT-5.2, GPT-5.5, and DeepSeek-V4, and benchmarks UltraFeedback, AlpacaEval, GSM8K, MATH, HumanEval, and MathArena. The results validate that (1) rational value risk is widespread; (2) value alignment can reduce, but cannot eliminate, it; (3) the risk is highly sensitive to inference-time reasoning strategy; and (4) longer reasoning improves rationality with diminishing returns. The code is at https://github.com/EVIEHub/LLM-Rationality.
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:2606.20621unread
PEAR: Permutation-Equivariant Adaptive Routing Multi-Agent Debate
Yang Feng, Ziwei Xu, Xia Hu, Fengxiang He · 2026-06-23
arXiv:2606. 20621v1 Announce Type: new Abstract: Multi-agent debate improves the reliability of large language models (LLMs) through iterative peer critiques.
Read next because PEAR: Permutation-Equivariant Adaptive Routing Multi-Agent Debate 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, eval, line, rate, position. Source: arxiv cs.AI (Artificial Intelligence).
arXiv:2606.20621v1 Announce Type: new Abstract: Multi-agent debate improves the reliability of large language models (LLMs) through iterative peer critiques. However, fixed topologies often introduce persistent positional biases, amplify unreliable agents, and cause high sensitivity to role assignments. We introduce \textit{Permutation-Equivariant Adaptive Routing Multi-Agent Debate (PEAR)}, an inference-time protocol that dynamically reconfigures communication roles and sparse topologies across consecutive debate rounds. By strategically switching agent-to-role assignments based on evolving agent states, PEAR prevents any agent from permanently occupying a privileged network position or distributes influence more evenly across the debate. We theoretically characterize PEAR as an equivariant sparse router: it preserves accuracy under agent relabeling while reducing routing complexity and improving generalization. Comprehensive empirical evaluations across four reasoning benchmarks and six diverse LLM backbones demonstrate PEAR significantly improves average accuracy over the strongest debate baselines. The code is at https://github.com/EVIEHub/PEAR.
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, benchmark.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.23515unread
FairBED: A Bayesian Experimental Design Approach to Gathering Fairer Data
Marcel Hedman, Emily Alger, Brieuc Lehmann, Chris Holmes, Tom Rainforth · 2026-06-23
arXiv:2606. 23515v1 Announce Type: new Abstract: Frameworks for ensuring fairness in machine learning typically focus on learning fair models from existing data.
Read next because FairBED: A Bayesian Experimental Design Approach to Gathering Fairer Data 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 "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: under, compare, trained, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.23515v1 Announce Type: new Abstract: Frameworks for ensuring fairness in machine learning typically focus on learning fair models from existing data. But this endeavor is often undermined by biases already present in that data. We therefore look to modify the data acquisition process itself to help gather fairer data that is inherently more suitable for training fair predictors. To this end, we introduce FairBED, which provides novel formulations for quantifying the fairness of datasets themselves based on the idea that fair datasets should be uninformative about sensitive attributes. We then use this to construct practical fairness-aware Bayesian experimental design (BED) objectives that maximize expected information gain about the target quantity of interest while minimizing expected information gain about sensitive attributes. We further derive a theoretical link between FairBED and demographic parity, and show empirically that models trained on data gathered using FairBED provide improved fairness-accuracy trade-offs compared to randomly acquired data and conventional BED.
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:2606.23477unread
Sublinearly Structured Deep Neural Networks Achieve Feature Learning Consistency for Compositional Functions
Sehwan Kim, Yan Sun, Faming Liang · 2026-06-23
arXiv:2606. 23477v1 Announce Type: new Abstract: Over the past decade, deep neural networks (DNNs) have achieved remarkable success on complex machine-learning tasks, yet the theoretical foundations of their performance remain incomplete.
Read next because Sublinearly Structured Deep Neural Networks Achieve Feature Learning Consistency for Compositional Functions 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, full, position, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.23477v1 Announce Type: new Abstract: Over the past decade, deep neural networks (DNNs) have achieved remarkable success on complex machine-learning tasks, yet the theoretical foundations of their performance remain incomplete. From a statistical viewpoint, a natural question is: can DNNs attain feature-learning and prediction consistency comparable to that of classical models? While a full characterization is open, we provide positive results for a broad subclass. We establish feature-learning consistency guarantees for sublinearly structured DNNs-architectures whose input/output dimensions and number of hidden neurons grow sublinearly with the sample size-when learning hierarchically compositional target functions. Importantly, this consistency still holds even in the conventional "over-parameterized" regime where the total number of parameters exceeds the number of training samples. Empirically, sublinearly structured DNNs match or surpass wide DNNs in prediction. A structural audit further indicates that widely used convolutional neural networks (CNNs), including AlexNet, VGGNet, ResNet, GoogLeNet, are sublinearly structured on their image classification benchmarks. We further prove that the sublinearly structured DNNs achieve universal approximation for hierarchically compositional functions in the large-sample limit. Moreover, images exhibit an inherent hierarchical, compositional structure. Taken together, these results explain, through a statistical lens, why many large-scale deep learning models succeed after adequate training on massive image datasets.
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 stat.ML (Machine Learning)arxiv:2606.22639unread
Statistical Inference for Misspecified Contextual Bandits
Yongyi Guo, Ziping Xu · 2026-06-23
arXiv:2606. 22639v1 Announce Type: new Abstract: Contextual bandit algorithms have transformed modern experimentation by enabling real-time adaptation for personalized treatment.
Read next because Statistical Inference for Misspecified Contextual Bandits 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, class, under, line, rate, project, without. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.22639v1 Announce Type: new Abstract: Contextual bandit algorithms have transformed modern experimentation by enabling real-time adaptation for personalized treatment. Yet these advantages create challenges for statistical inference due to adaptivity. We study inference with contextual-bandit data without assuming a well-specified outcome model. In this setting, we show a previously overlooked issue: standard algorithms such as LinUCB may fail to stabilize under misspecified working models, leading to non-Gaussian estimator behavior and invalid inference. This issue is practically important, as misspecified working models -- such as approximations of complex dynamical systems -- are often employed by online agents in real-world adaptive experiments to balance reward, computational tractability, and robustness. We develop an inverse-probability-weighted Z-estimation framework for a broad class of marginal moment targets, including projection parameters, structural parameters with noisy contexts, and off-policy values. We identify a stability condition tailored to this framework, scaled inverse-propensity convergence, under which the IPW-Z estimator is consistent and asymptotically normal with a consistent sandwich variance estimator. We further establish sufficient conditions for scaled inverse-propensity convergence for several policy classes, including multi-armed bandit algorithms and smooth contextual allocation policies. Simulations and a HeartSteps V1 real-data-calibrated application show reliable coverage and competitive performance across multiple targets. Overall, our results highlight the importance of stability-aware adaptive design for valid post-experiment inference.
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 stat.ML (Machine Learning)arxiv:2606.22521unread
Robust Diffusion Models via Divergence-Induced Weighted Denoising
Lei Li, Yuexiao Dong · 2026-06-23
arXiv:2606. 22521v1 Announce Type: new Abstract: We show that replacing the standard MSE denoising loss in diffusion models with a nonlinear transformation induced by an f-divergence yields a simple robust training surrogate that empirically improves performance under data contamination, with small additional computational overhead.
Read next because Robust Diffusion Models via Divergence-Induced Weighted Denoising 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, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.22521v1 Announce Type: new Abstract: We show that replacing the standard MSE denoising loss in diffusion models with a nonlinear transformation induced by an f-divergence yields a simple robust training surrogate that empirically improves performance under data contamination, with small additional computational overhead. The theoretical foundation rests on a local divergence construction: under the Gaussian reverse-kernel structure of DDPM, each per-step likelihood ratio follows a lognormal distribution parameterized by a scalar mismatch, so the conditional f-divergence at each step reduces to a one-dimensional function of the denoising error. Summing these local divergences yields a training objective that unifies diffusion training as divergence induced weighted denoising, where the derivative of the induced divergence acts as a residual-space influence weight that controls the contribution of each sample. Bounded-influence divergences (Hellinger, negative exponential) suppress large error samples, with Hellinger yielding an explicit exponential weight, connecting the framework to robust M-estimation. Empirically, on CIFAR-10 under 30% contamination, NED reduces FID from 93.0 (KL) to 77.5, while also outperforming standard robust losses such as Huber and clipped MSE.
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 negative.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.22346unread
Flow Annealing Posterior Sampling for Function-Space Regression and Inverse Problems
Yaozhong Shi, Zachary E. Ross, Yisong Yue · 2026-06-23
arXiv:2606. 22346v1 Announce Type: new Abstract: Principled regression for stochastic processes is a long-standing challenge with deep connections to scientific inverse problems.
Read next because Flow Annealing Posterior Sampling for Function-Space Regression and Inverse Problems 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, correct, eval, line, rate, trained, test. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.22346v1 Announce Type: new Abstract: Principled regression for stochastic processes is a long-standing challenge with deep connections to scientific inverse problems. We introduce Flow Annealing Posterior Sampling (FAPS), to our knowledge the first function-space posterior sampling framework that unifies stochastic-process regression and PDE inverse problems. Built on pretrained function-space flow-matching priors, FAPS enables likelihood-guided posterior inference from sparse and noisy observations, supports variable query discretizations, and avoids explicit prior-density evaluation. Its Langevin correction uses a low-rank covariance preconditioner to exploit dominant function-space correlations across discretizations. Across Gaussian and non-Gaussian stochastic-process regression benchmarks and diverse PDE inverse problems, FAPS produces coherent posterior samples with accurate uncertainty quantification, significantly outperforming existing functional regression baselines and achieving competitive or better PDE noisy inverse performance than diffusion-based posterior samplers while reducing test-time sampling cost.
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, benchmark.
- score 100arxiv stat.ML (Machine Learning)arxiv:2606.20880unread
Adversarial observations in probabilistic State-Space Models for robust Reinforcement Learning
M. Santos-Pascual, D. R\'ios Insua · 2026-06-23
arXiv:2606. 20880v1 Announce Type: new Abstract: Decision-making under partial or adversarial observability requires accurate inference of the environment's latent state and its associated uncertainty.
Read next because Adversarial observations in probabilistic State-Space Models for robust 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 "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, under, line, rate, model. Source: arxiv stat.ML (Machine Learning).
arXiv:2606.20880v1 Announce Type: new Abstract: Decision-making under partial or adversarial observability requires accurate inference of the environment's latent state and its associated uncertainty. This work analyses adversarial attacks on linear probabilistic state-space models, commonly integrated within reinforcement learning architectures, where the attacker alters observations under likelihood constraints that ensure the perturbations remains consistent. We analyze how such adversarial yet realistic observation shifts influence the latent state and influence policy decisions. This perspective provides a principled pathway toward building more robust reinforcement learning systems, with direct relevance to safety-critical domains such as robotics, where reliable operation under sensor noise, partial failures, and adversarial conditions is essential.
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, failures, adversarial.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21592unread
Enhancing Stateful Detection of Adversarial Attacks with Soft-labels' Temporality and Robust Similarity Approximations
De Zhang Lee, Han Fang, Ee-Chien Chang · 2026-06-23
arXiv:2606. 21592v1 Announce Type: new Abstract: Stateful Detection (SD) mitigates adversarial attacks by determining whether a sequence of queries contains queries from a black-box adversary.
Read next because Enhancing Stateful Detection of Adversarial Attacks with Soft-labels' Temporality and Robust Similarity Approximations 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, soft, rate, implement. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21592v1 Announce Type: new Abstract: Stateful Detection (SD) mitigates adversarial attacks by determining whether a sequence of queries contains queries from a black-box adversary. Recent works, such as Blacklight and PIHA utilize query similarity to detect such queries. In this paper, we observe that temporal information, in particular, the temporal correlation of the classification soft labels, is a prominent characteristic of adversarial attacks and can be leveraged to reduce false positive rates. Moreover, we point out a potential vulnerability in SD implementation. Many SD systems identify similar queries according to some implicit, computationally expensive metric. To improve efficiency, these systems often adopt an approximate similarity function as substitute. This discrepancy could be exploited by crafting queries that appear dissimilar under the approximation but are close in the intended metric, thereby evading detection. We refer to this as an ``adversarial attack'' on the approximation function, and demonstrate it through a lightweight attack on Blacklight's similarity function. Based on the above observations, we propose a two-phase approach. The first phase identifies subsequences of queries with high similarity, incorporating randomness to prevent the aforementioned ``adversarial attacks''. The second phase analyzes temporal correlation of the soft-labels to further validate the presence of the adversary's queries. Experimental results show that the framework detects adversarial queries generated by Boundary Attack, HSJA, SimBA, Square Attack with true positive rate (TPR) reaching 1.00, while maintaining a false positive rate (FPR) of at most 0.06. Additionally, the method is robust against OARS which is an adaptive attack.
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.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21397unread
Evaluating LLMs for Real-World Web Vulnerability Detection
Sebastian Neef, Luca Jungnickel, Antonio Benjamin Buchholz, Valene Spence, Vicente Birke Gonzalez · 2026-06-23
arXiv:2606. 21397v1 Announce Type: new Abstract: Large Language Models (LLMs) have emerged as a promising tool for automated vulnerability detection, yet their effectiveness on web-specific vulnerabilities remains to be explored.
Read next because Evaluating LLMs for Real-World Web Vulnerability 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, word, rect, correct, eval, line, rate, full. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21397v1 Announce Type: new Abstract: Large Language Models (LLMs) have emerged as a promising tool for automated vulnerability detection, yet their effectiveness on web-specific vulnerabilities remains to be explored. This work benchmarks six frontier (Claude Opus 4.6, Codex GPT-5.4, Gemini 3.1-pro-preview) and open-weight models (Qwen 3.5, Qwen 3 Coder Next, MiniMax M2.5) on their ability to detect real-world web vulnerabilities using static analysis in WordPress plugins, including SQL injection, stored cross-site scripting, path traversal, and remote code execution. Using five prompt designs of varying structure, scope, and complexity across three experiment iterations, we aim to answer how model and prompt choice affects vulnerability detection. Our results show that all models are capable of detecting valid security issues, but the detection rate varies depending on the model and prompt. For example, Claude Opus 4.6 achieved the highest web vulnerability detection rate (63%), while open-weight MiniMax M2.5 performs on par with other frontier models (48%), and self-hosted Qwen 3.5 only achieved 35%. We show that scoped prompts that narrow the vulnerability scope outperform open-ended ones, whereas the prompt complexity has little impact. Surprisingly, no model achieved full reporting consistency across three experiment iterations, with some as low as 50%. Our experiments demonstrate the opportunities and limits of LLM-based vulnerability detection, as no model correctly identified one baseline vulnerability in one of the plugins. Additionally, we derive practical lessons learned for security practitioners and publish all code and data to support future 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 benchmark.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21353unread
Beyond Classification Accuracy: An Exploration-Range Evaluation of Adaptive Crawling for Fake Shopping Sites
K. Karasawa, K. Takeshige, S. Matsugaya, M. Shimamura, M. Hashimoto · 2026-06-23
arXiv:2606. 21353v1 Announce Type: new Abstract: In recent years, fake shopping sites targeting Japanese users have appeared in the top results of search engines through SEO poisoning, causing increasing damage.
Read next because Beyond Classification Accuracy: An Exploration-Range Evaluation of Adaptive Crawling for Fake Shopping Sites 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, word, class, eval, line, rate, lora. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21353v1 Announce Type: new Abstract: In recent years, fake shopping sites targeting Japanese users have appeared in the top results of search engines through SEO poisoning, causing increasing damage. Conventional collection methods rely on fixed keywords and cannot keep up with evolving attack campaigns, delaying the discovery of new sites. We propose a closed-loop crawler that incorporates the page-level outputs of a fake-site classifier (fastText+LightGBM) into the search queries of the next cycle. Search queries are generated by a seed-compound strategy that combines characteristic words extracted from positive pages with seed words from the fake-shopping context (e.g., ``deep discount,'' ``official''). To complement evaluations that tend to focus on classifier accuracy, we also introduce per-cycle new-host counts and cumulative unique-host counts as exploration-range metrics. In a comparative experiment ($n=3$ for the proposed method, $n=2$ for the baseline), the fixed-keyword baseline yielded zero new-host acquisition from cycle 2 onward, indicating complete stagnation, whereas the proposed method continued to discover new hosts and, at cycle 3, achieved a cumulative unique-host count approximately 7.6 times that of the baseline on average.
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:2606.21282unread
Differential Zonotopes for Verifying Global Robustness of DNNs
Anagha Athavale, Samuel Teuber, Matteo Maffei, Ezio Bartocci, Dejan Nickovic, Georg Weissenbacher · 2026-06-23
arXiv:2606. 21282v1 Announce Type: new Abstract: The robustness of deep neural networks (DNNs) is critical in security-sensitive applications, where small input perturbations should not alter model predictions.
Read next because Differential Zonotopes for Verifying Global Robustness of DNNs 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, eval, implement, propagate, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21282v1 Announce Type: new Abstract: The robustness of deep neural networks (DNNs) is critical in security-sensitive applications, where small input perturbations should not alter model predictions. This property is commonly formalized as local or global robustness: the former considers perturbations around a single input, while the latter -- strictly stronger -- quantifies over all input pairs. While local robustness can be expressed as a safety property, global robustness is a 2-safety property, making it substantially more challenging to verify. We present a novel static analysis technique for verifying the global robustness of DNNs. Our approach is based on differential halo zonotopes, a new abstract domain that extends zonotopes to jointly propagate pairs of perturbed inputs in lock-step while tightly bounding their divergence. In addition, we introduce a symmetric variant of confidence-based global robustness that disregards perturbations leading to differing but low-confidence predictions. This relaxation yields a practically meaningful notion of robustness that applies to a broader class of networks. We implement our approach in a new tool, called TwoSafe, and evaluate it on standard DNN verification benchmarks, including widely deployed models. Our results show that TwoSafe significantly outperforms the state of the art in both precision and scalability, enabling the verification of networks an order of magnitude larger than those handled by prior techniques.
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.CR (Cryptography and Security)arxiv:2606.21240unread
DIPBox: A Multi-scale Testing Framework for Tracking Dataset Regeneration
Tian Dong, Yan Meng, Shaofeng Li, Guoxing Chen, Yuling Chen, Zhen Liu, Haojin Zhu, Hao Chen · 2026-06-23
arXiv:2606. 21240v1 Announce Type: new Abstract: Training datasets have tremendous proprietary value and are vulnerable to unauthorized copying.
Read next because DIPBox: A Multi-scale Testing Framework for Tracking Dataset Regeneration 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, rate, full, test, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21240v1 Announce Type: new Abstract: Training datasets have tremendous proprietary value and are vulnerable to unauthorized copying. Existing defenses mainly focus on tracking individual data points, but pay little attention to the threat of dataset regeneration. Through a measurement study of public tumor datasets, we identify substantial real-world partial-dataset replication, raising concerns about potential license noncompliance. To counter the challenge of tracking previously unknown adversarial regeneration, our key insight is that regeneration that preserves model utility inevitably preserves measurable signals across multiple feature scales. We categorize these dataset features into sample-, set-, and distribution-level features and design four similarity metrics to accurately identify regeneration. Based on these metrics, we develop DIPBox, which to our knowledge is the first testing framework that tracks regeneration suspects via multi-scale similarity testing across a spectrum of defender access settings, from limited to full information. We further provide a learning-theoretic analysis that justifies these multi-scale metrics and formalizes an inherent utility--divergence trade-off, implying fundamental limits on evasive regeneration. Extensive experiments on 16 vision and text base datasets, 320 regenerated datasets, and 590 derived models validate that DIPBox outperforms previous solutions while characterizing its robustness and limits under three adaptive attacks.
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, adversarial.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21210unread
Impact Analysis of Speech Representation Learning Models for Acoustic Side-Channel Attack
Nitin Choudhury, Vikrant Vikram Pratap Maurya, Arun Balaji Budhuru, Orchid Chetia Phukan · 2026-06-23
arXiv:2606. 21210v1 Announce Type: new Abstract: Acoustic side-channel attacks (ASCA) on keyboards have gained increasing attention, yet impact of speech representation learning models in ASCA remains unexplored.
Read next because Impact Analysis of Speech Representation Learning Models for Acoustic Side-Channel Attack 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, line, full, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21210v1 Announce Type: new Abstract: Acoustic side-channel attacks (ASCA) on keyboards have gained increasing attention, yet impact of speech representation learning models in ASCA remains unexplored. Addressing this, we introduce KEYAC, a dataset designed to analyze representation generalization for ASCA under both standard and VoIP codec settings. On KEYAC, we evaluate six representation learning models under zero-shot and partial fine-tuning settings using fully connected and convolutional networks. Results show that while partial fine-tuning improves performance, models struggle to generalize across VoIP codecs. We hypothesize this limitation stems from inadequate modeling of nonlinear feature interactions in conventional fine-tuning architectures. To address this, we employ Kolmogorov-Arnold Networks (KAN) for fine-tuning. Empirical results show that KAN-based fine-tuning consistently outperforms the baselines and establishes a new state-of-the-art on KEYAC.
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.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21077unread
OTTER: A Red-Teaming System for Toxicity-Evading Jailbreak Prompt Optimization
Jerry Wang, Hsin-Ling Hsu, Yi-Cheng Lai, Nai-Chia Chen, Fang Yu · 2026-06-23
arXiv:2606. 21077v1 Announce Type: new Abstract: Production LLMs increasingly rely on toxicity-based moderation filters as a primary defense, assuming that harmful intent correlates with toxic surface wording.
Read next because OTTER: A Red-Teaming System for Toxicity-Evading Jailbreak Prompt Optimization 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: word, class, latin, rect, eval, token, model. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21077v1 Announce Type: new Abstract: Production LLMs increasingly rely on toxicity-based moderation filters as a primary defense, assuming that harmful intent correlates with toxic surface wording. We show this assumption is fundamentally brittle: surface toxicity and adversarial intent can be decoupled by replacing as few as five tokens. We present OTTER (Obfuscated Toxicity-Evading Token Evolution for Rewriting), a black-box red-teaming framework requiring only standard API access, directly targeting the practical constraints of industry security audits. Evaluated on 457 AdvBench prompts across four GPT models, OTTER raises average ASR from 7.0% to 84.0%. We further provide the first quantitative analysis of the toxicity--bypass relationship and a per-category breakdown, translating our findings into actionable recommendations for classifier hardening in production deployments.
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.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.21071unread
Local LLM Agents as Vulnerable Runtimes:A Source-Code Audit of the Agent Runtime Layer
Zhengsong Zhang, Zongze Li, Jiawei Guo, Haipeng Cai · 2026-06-23
arXiv:2606. 21071v1 Announce Type: new Abstract: Local LLM agents such as OpenClaw and Nanobot run on end-user machines and act on host resources - the shell, filesystem, browser, stored credentials, and messaging applications - through natural-language goals.
Read next because Local LLM Agents as Vulnerable Runtimes:A Source-Code Audit of the Agent Runtime Layer 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, soft, eval, source, line, implement, test, language. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21071v1 Announce Type: new Abstract: Local LLM agents such as OpenClaw and Nanobot run on end-user machines and act on host resources - the shell, filesystem, browser, stored credentials, and messaging applications - through natural-language goals. These agents have become privileged software runtimes that mediate between user intent, model outputs, and host-level actions. Existing research characterizes the landscape through prompt injection, malicious skills, marketplace risks, or black-box evaluation of agents. But the implementation layer that performs this mediation, the prompt builder, parser, tool dispatcher, skill loader, memory writer, network client, and permission gate, has remained an unexamined safety boundary. To our knowledge, no prior work has examined the agent's source tree to audit these components for implementation-level security weaknesses. We present CLAWAUDIT, a static auditing framework for measuring vulnerability exposure in local LLM agent runtimes. CLAWAUDIT derives a five-category vulnerability taxonomy from STRIDE and develops custom static-analysis rules that target agent-specific patterns absent from established rule sets for vulnerability analysis. We instantiate the taxonomy in two backends, 47 Semgrep YAML rules and 30 CodeQL queries, and evaluate on OPENCLAWBENCH, a benchmark of 446 source-code-level advisories from the OpenClaw repository and split temporally into 229 rule-derivation (train) and 217 held-out (test) advisories. On the held-out test, CLAWAUDIT raises Semgrep recall from 21.7% (Pro baseline) to 66.8%, and CodeQL recall from 13.8% (security-extended) to 75.1%. Train/test gaps remain within 4 percentage points for all four configurations, indicating that the rules generalize to vulnerabilities unseen during rule writing. A preliminary live-code audit shows that these recall-oriented rules require manual triage, motivating semantic filtering before production deployment.
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:2606.21059unread
DEFENGRAPH: Knowledge Graph-Enhanced LLMs for Blue Team Cyber Defense
Zhen Wang, Kristen Moore, Qin Wang, Guangsheng Yu, Minjune Kim, Diksha Goel, Gang Li, Ahmed Ibrahim, Ahmad Mohsin, Helge Janicke · 2026-06-23
arXiv:2606. 21059v1 Announce Type: new Abstract: Large Language Models (LLMs) show promise for supporting decision-making in cybersecurity, but their reliability in high-stakes, time-evolving environments remains limited due to hallucinations, poor temporal reasoning, and shallow grounding in system context.
Read next because DEFENGRAPH: Knowledge Graph-Enhanced LLMs for Blue Team Cyber Defense 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, latin, rect, correct, eval, assistant, line, rate. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21059v1 Announce Type: new Abstract: Large Language Models (LLMs) show promise for supporting decision-making in cybersecurity, but their reliability in high-stakes, time-evolving environments remains limited due to hallucinations, poor temporal reasoning, and shallow grounding in system context. We introduce DEFENGRAPH, an LLM-driven assistant designed to support human defenders during cybersecurity incidents. DEFENGRAPH improves contextual reasoning by integrating a dual-layer Static-Dynamic Knowledge Graph (KG) with graph-based path retrieval, LLM-driven contextual filtering, and reasoning-based re-ranking. The framework grounds LLM outputs in both long-term domain knowledge and evolving event context, enabling faithful and temporally aware decision support. We evaluate DEFENGRAPH in a cyber defense setting using knowledge graphs constructed from heterogeneous security artifacts, including SIEM alerts, system topology, attacker behaviors, and prior defensive actions. The evaluation uses data collected during live Red vs. Blue team cyber range exercises simulating attacks on critical infrastructure, which generate realistic and noisy datasets reflecting real-world defender workflows and system dynamics. Evaluations across four prevalent LLMs show that DEFENGRAPH sets a new state-of-the-art: on GPT-4o it boosts reasoning-recall from 61.45\% to 73.49\% and ticket-action recall from 52.17% to 72.46% (precision 24.49\% to 29.24\%), with similar gains on LLaMA-3 (46.99\% to 61.45\%), DeepSeek-R1 (45.78\% to 56.63\%) and QWen-3 (51.81\% to 59.04\%), while surfacing up to 50 correct defense actions versus 36 for the next best baseline and holding fault rates steady.
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:2606.21037unread
Honeyquest for LLMs: Rethinking Cyber Deception for AI Attackers
Kerri Prinos, Lilianne Brush, Cameron Denton · 2026-06-23
arXiv:2606. 21037v1 Announce Type: new Abstract: The empirical foundation of cyber deception relies on human-centered hypotheses, but the rapid emergence of autonomous, AI-enabled attackers challenges whether this foundation transfers to AI agents.
Read next because Honeyquest for LLMs: Rethinking Cyber Deception for AI Attackers 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, line, rate, full, model, absent. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21037v1 Announce Type: new Abstract: The empirical foundation of cyber deception relies on human-centered hypotheses, but the rapid emergence of autonomous, AI-enabled attackers challenges whether this foundation transfers to AI agents. To address this, we introduce an automated evaluation framework adapted from the Honeyquest instrument to assess LLM attacker judgment at scale. Our 21-LLM cohort spanned 10 providers, diverse architectures and specializations, open- and closed-weight models, and parameter scales from 8B to over 1T. We evaluated the performance of this LLM cohort (yielding 10,962 responses) against the 47-participant human baseline across an identical set of 174 reconnaissance queries. Our empirical evaluation reveals three key findings that establish LLMs as a distinct attacker class: (1) every model in our cohort falls for deceptive traps at a significantly higher rate than human attackers; (2) the defensive attention-diversion effect observed in humans is statistically absent in our LLM cohort; and (3) a critical recognition-action gap, where LLMs successfully articulate trap recognition in their reasoning but exploit the deceptive elements anyway 73.4\% of the time. Across the 21 models, trap recognition in reasoning text did not predict fell-for-trap behavior (Spearman $r = +0.08$, $p = 0.73$). Ultimately, these findings demonstrate that human-centered deception hypotheses do not reliably transfer to AI attackers, highlighting the critical need for new research into AI-native active defense frameworks.
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:2606.21016unread
Quantifying the Impact of Stealthy BLE Spam & Flooding Attacks on IoT Environments
Usman Rauf, Adalynn Martinez, Fadi Mohsen · 2026-06-23
arXiv:2606. 21016v1 Announce Type: new Abstract: The energy-efficient design of the BLE protocol, emphasis on rapid, and userfriendly discovery, making it an ideal choice for IoMTs, specifically, military field medical systems, and battlefield wearable sensors.
Read next because Quantifying the Impact of Stealthy BLE Spam & Flooding Attacks on IoT Environments 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, under, source, rate, test. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.21016v1 Announce Type: new Abstract: The energy-efficient design of the BLE protocol, emphasis on rapid, and userfriendly discovery, making it an ideal choice for IoMTs, specifically, military field medical systems, and battlefield wearable sensors. Especially in active conflict zones, when static medical facilities are vulnerable and often targeted, limiting their viability for sustained care delivery. This rapid deployment, and ease of management comes at the cost of expanded attack surface, i.e., BLE flooding attacks. During such attacks, adversaries flood advertisement channels with unauthorized connection or advertising requests to exhaust nearby device resources and disrupt legitimate communication, sometimes culminating in denial-of-service conditions. A first public proof-of-concept of such attacks, using a Raspberry Pi has since been adapted to commodity platforms (e.g., Flipper Zero, HackRF, Android), lowering the barrier to attack. In contested environments, such platforms are directly relevant to adversarial RF jamming and spoofing operations, where low-cost, portable devices can induce disproportionate disruption in dense wireless ecosystems. In this work, we develop a quantitative foundation for understanding the impact of such attacks and propose a practical deterrence strategy based on agility to raise the cost of such attacks.
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:2606.20922unread
Think Twice Before You Act: Protecting LLM Agents Against Tool Description Poisoning via Isolated Planning
Shanghao Shi, Xiao Wang, Chaoyu Zhang, Hao Li, Wenjing Lou, Thomas Hou, Yevgeniy Vorobeychik, Chongjie Zhang, Ning Zhang · 2026-06-23
arXiv:2606. 20922v1 Announce Type: new Abstract: The integration of external tools has substantially expanded the capabilities of large language model (LLM) agents, but it also introduces new attack surfaces beyond prompt injection.
Read next because Think Twice Before You Act: Protecting LLM Agents Against Tool Description Poisoning via Isolated Planning 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, language, model, never. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.20922v1 Announce Type: new Abstract: The integration of external tools has substantially expanded the capabilities of large language model (LLM) agents, but it also introduces new attack surfaces beyond prompt injection. In particular, cross-tool description poisoning can manipulate planner-visible tool metadata to steer an agent's trajectory, even if the poisoned tool itself is never chosen. To understand the effectiveness of existing defenses against this emerging threat, we first evaluate several prompt-injection defenses and find that they transfer poorly to cross-tool description poisoning. A key observation is that poisoned descriptions persist in the planning context across steps, enabling continuous influence over subsequent tool choices. Building on this insight, we propose Tool-Guard, a novel system-level defense based on a new concept called isolated planning, in which tool invocations that are detected as misaligned or suspicious cause the corresponding tool to be placed in a quarantined list (the influenced list), breaking further influence from poisoned descriptions. With this influence isolated, the tool can continue to be used to support the task, enabling a robust defense that preserves legitimate tool utility. Experiments on the AgentDojo and ASB benchmarks show that Tool-Guard substantially reduces attack success while maintaining high task utility. Our code is available at https://github.com/shishishi123/Tool-Guard.
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.CR (Cryptography and Security)arxiv:2606.20868unread
Can LLMs Reason About Brand Ownership? An Empirical Study of Domain Attribution Intelligence
Fathima Mashood, Mohamed Nabeel · 2026-06-23
arXiv:2606. 20868v1 Announce Type: new Abstract: When a new domain resembling a popular brand appears, defenders face a fundamental ambiguity: it may be an attacker-created squatting site for phishing, or it may be a domain the brand itself registered, either defensively, to block attackers, or legitimately, for a new product or service launch.
Read next because Can LLMs Reason About Brand Ownership? An Empirical Study of Domain Attribution Intelligence 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, rect, correct, eval, line, without, alone. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.20868v1 Announce Type: new Abstract: When a new domain resembling a popular brand appears, defenders face a fundamental ambiguity: it may be an attacker-created squatting site for phishing, or it may be a domain the brand itself registered, either defensively, to block attackers, or legitimately, for a new product or service launch. Incorrectly flagging a brand-owned domain as malicious produces a false positive that harms end users and damages the brand's reputation. Resolving this ambiguity requires brand intelligence: the ability to determine, at scale, whether a given domain belongs to a brand. Large language models (LLMs), with their broad knowledge of brand domain relationships, offer a promising zero configuration approach to this problem, but their reliability for brand intelligence tasks remains unknown. We present the first systematic empirical evaluation of LLM brand intelligence across three tasks: domain enumeration (Q1), open ended brand attribution (Q2), and binary ownership classification (Q3). We evaluate four models, Gemini 2.5 Flash, Gemini 3.5 Flash, Claude Sonnet 4.5, and Claude Sonnet 4.6, across four retrieval settings (in context, web search, WHOIS lookup, and combined) on 36 of the most phished brands. Our results reveal a stark dichotomy: models achieve up to 82% precision enumerating brand domains from memory alone, yet fail at ownership verification without external tools, with macro F1 at most 0.37 in ICL mode. WHOIS augmentation lifts Q3 macro F1 by up to 0.65 points, yielding near perfect precision (<= 0.99), dramatically reducing the false positive risk for defenders. We provide concrete recommendations for deploying LLMs in brand protection pipelines.
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:2606.20864unread
Synthetic Network Packet Generation through Statistical Learning and Genetic Algorithms
Mayank Raj, Nathaniel D. Bastian, Lance Fiondella, Gokhan Kul · 2026-06-23
arXiv:2606. 20864v1 Announce Type: new Abstract: Developing robust intrusion detection systems (IDS) for IoT environments requires large, labeled datasets capturing realistic traffic distributions across both benign and malicious activity.
Read next because Synthetic Network Packet Generation through Statistical Learning and Genetic Algorithms 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, rect, under, distributional, eval, line, rate. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.20864v1 Announce Type: new Abstract: Developing robust intrusion detection systems (IDS) for IoT environments requires large, labeled datasets capturing realistic traffic distributions across both benign and malicious activity. Existing public datasets suffer from fixed activity distributions and extreme class imbalance, while deep generative models (GANs, VAEs) provide no mechanism to enforce that synthetic packets remain within physically valid feature ranges. This paper proposes and compares two constraint-enforcing approaches for synthetic IoT network packet generation: (i) a statistical learning method combining PCA-based latent space sampling with dual One-Class SVM (OCSVM) and Isolation Forest (IF) boundary enforcement, and (ii) a genetic algorithm (GA) method that treats packet generation as a multi-objective optimization problem with explicit fitness criteria for anomaly model acceptance and distributional fidelity. Both methods embed hard validity constraints -- dual anomaly-detection gating, feature-range clamping, and independent validation -- directly into the synthesis pipeline. Evaluation on the complete ACI IoT 2023 dataset (1,231,411 packets, 12 attack categories, class imbalance up to 175,805:1) demonstrates that both methods achieve PASS status across all categories under independently trained validators with a 30% anomaly rate threshold: the statistical method attains 1.20% average anomaly rate with ~1,091 packets/s throughput, while the GA attains 0.62% average anomaly rate with organic per-class variance (0.00%-2.50%) at ~5.7 packets/s. Both methods successfully amplify the 5-sample ARP Spoofing category by 200x to 1,000 validated packets. The ~190:1 throughput ratio between methods, combined with their complementary quality profiles, provides evidence-based selection criteria for deployment contexts ranging from rapid dataset augmentation to adversarial robustness testing.
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, adversarial, evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.20835unread
PromptMark: A Prompt-Guided Iterative-Feedback Framework for Source Code Watermarking
Istiaq Ahmed Fahad, Mridha Md. Nafis Fuad, Kazi Sakib · 2026-06-23
arXiv:2606. 20835v1 Announce Type: new Abstract: Watermarking has become a crucial technique for ensuring provenance and accountability in AI-generated source code.
Read next because PromptMark: A Prompt-Guided Iterative-Feedback Framework for Source Code Watermarking 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, under, correct, eval, source, line. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.20835v1 Announce Type: new Abstract: Watermarking has become a crucial technique for ensuring provenance and accountability in AI-generated source code. As large language models (LLMs) are increasingly integrated into development workflows, reliable attribution remains challenging. In practice, most developers rely on commercial LLM APIs operating under black-box constraints, making existing approaches that require access to the decoding process less feasible for real-world integration. To address this limitation, we propose PromptMark, a black-box, prompt-guided watermarking framework that embeds invisible yet statistically detectable signals into generated code via structured input instructions. The method steers models toward subtle identifier and comment naming patterns while preserving the functional correctness and structural integrity of the generated code. Detection is performed using statistical tests designed to remain reliable across varying code lengths and model outputs. The embedding is further refined through an iterative feedback loop, where prompts are updated based on watermark detection scores. Experiments on the MBPP and HumanEval benchmarks show that PromptMark consistently achieves strong watermark detectability while maintaining high code correctness, outperforming baseline approaches.
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, benchmark.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.20786unread
Memory-Centric Computing: Security Benefits and Challenges of Processing-in-DRAM
Ismail Emir Yuksel, F. Nisa Bostanci, Ataberk Olgun, Onur Mutlu · 2026-06-23
arXiv:2606. 20786v1 Announce Type: new Abstract: Today's computing systems are processor-centric: they require frequent data movement between processing elements (e.
Read next because Memory-Centric Computing: Security Benefits and Challenges of Processing-in-DRAM 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, capability. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.20786v1 Announce Type: new Abstract: Today's computing systems are processor-centric: they require frequent data movement between processing elements (e.g., CPU) and main memory (DRAM), leading to significant inefficiencies in performance and energy consumption. Memory-centric computing instead moves computation to the data, enabling computation capability in and near all places where data is generated and stored, and greatly reducing the performance and energy overheads of data access and data movement. This shift from a processor-centric to a memory-centric paradigm has important and underexplored consequences for system security. Turning memory from a dumb, inactive store into an active computing substrate introduces benefits as well as challenges for system security: it can provide new in-memory security primitives and also reduce data exposure, but it can also expose new attack surfaces. This work discusses the security benefits and challenges of memory-centric computing, specifically Processing-in-DRAM (PiD), a paradigm where the operational characteristics of a DRAM chip are exploited and enhanced to perform computation on data stored in DRAM. Specifically, we describe 1) new state-of-the-art DRAM-based true random number generators that provide up to 16.05 Gb/s throughput and physical unclonable functions with 5.75% lower evaluation latency than the prior state-of-the-art, both on real DRAM chips and 2) two key security challenges of PiD: amplified DRAM read disturbance (e.g., 158x reduction in the minimum number of DRAM accesses required to induce the first bitflip) and high throughput memory timing channels (e.g., a communication throughput of 14.8Mb/s). We believe it is time to design, use, and program DRAM, and in general memory, not as an inactive storage substrate, but as a combined computation, storage, and security substrate, where computational capability, storage density, and security are all key goals.
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:2606.20760unread
Privacy-Preserving Compliance on Public Ledgers via Selective Disclosure Authorization Schemes
Supriya Khadka, Sanchari Das · 2026-06-23
arXiv:2606. 20760v1 Announce Type: new Abstract: Public distributed ledgers enforce integrity through radical transparency, creating tension with data minimization principles required for regulatory compliance.
Read next because Privacy-Preserving Compliance on Public Ledgers via Selective Disclosure Authorization Schemes 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, implement, control, without, binding, chain. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.20760v1 Announce Type: new Abstract: Public distributed ledgers enforce integrity through radical transparency, creating tension with data minimization principles required for regulatory compliance. While Zero-Knowledge Proofs (ZKPs) offer a theoretical privacy solution, existing constructions often overlook adversarial constraints in smart contract environments. Specifically, the asynchronous decoupling of off-chain proof generation from on-chain submission introduces front-running and proof-reuse risks in public mempools. In this work, we formalize Selective Disclosure Authorization Schemes (SDAS), a cryptographic primitive for granular and revocable compliance checks on public ledgers without revealing the underlying witness. We define a security model for SDAS, introducing Ledger-Bound Attribute Unlinkability and Context-Aware Sender Binding to capture how valid proofs remain bound to their intended authorization context. To validate sender binding, we present ZK-Compliance, an Ethereum-based instantiation that operationalizes a user-controlled "Grant, Verify, Revoke" lifecycle. We implement the sender-binding component using a 14-constraint Circom circuit that anchors the zero-knowledge proof to the executing on-chain sender address. Our Sepolia evaluation confirms practical viability: browser-based proof generation executes in under 200 ms, and on-chain verification costs 240,512 gas, neutralizing proof reuse by different callers while preserving strict attribute privacy.
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, evaluation.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.20746unread
Amplify, Don't Create: Temporal Accumulation for Slow-Burn Prompt Injection
J Alex Corll · 2026-06-23
arXiv:2606. 20746v1 Announce Type: new Abstract: Most prompt-injection detectors score a single event or message.
Read next because Amplify, Don't Create: Temporal Accumulation for Slow-Burn Prompt Injection 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, directive, under, eval, rate, control, does, test. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.20746v1 Announce Type: new Abstract: Most prompt-injection detectors score a single event or message. Control-plane attacks against tool-using agents can instead distribute weak directives across a trajectory while keeping each event below threshold. We test whether a proxy-side temporal accumulator recovers this slow-burn signal by reducing frozen per-event scores to peak and CUSUM persistence statistics. To avoid circularity, grafts are generated against a held-out autoregressive cloaking target and then re-scored under a detector of record: a frozen char-ngram SVM plus an embedding-contrastive head. Only floor-met grafts bound to executed action edges and still sub-threshold under the detector of record enter the slow-burn endpoint. This is a boundary result, not a deployable detector. On concentrated attacks, trajectory-level accumulation beats the per-event foil under a clustered bootstrap (gap +0.092, 95% CI [+0.025, +0.155]), while persistence and peak are statistically tied. On git repo-exfil, density-four floor-met sub-threshold grafts add persistence mass that matched benign shams do not (persistence-delta AUC 0.708 over four attack survivors and six benign shams), while the matched peak-delta control does not separate attack from sham (AUC 0.417), localizing the effect to accumulated persistence rather than a single hot graft. The effect fails on broader clean-path actions (persistence-delta AUC 0.167), where the detector assigns attack and benign actions indistinguishable per-event scores, leaving no margin for CUSUM to bank. Independent powering is blocked by only three to four independent tasks. Temporal accumulation is therefore a narrow-band margin amplifier: it can bank elevated sub-threshold signal but cannot create margin where the per-event detector has none. As byproducts, we contribute a pseudo-replication warning and an independence-audit standard for agent-benchmark evaluation.
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, benchmark.
- score 100arxiv cs.CR (Cryptography and Security)arxiv:2606.20668unread
BELLS-O: Evaluating the Operational Trade-offs of LLM Supervision Systems
Leonhard Waibl, Felix Michalak, Hadrien Mariaccia · 2026-06-23
arXiv:2606. 20668v1 Announce Type: new Abstract: LLM supervision systems, namely input/output moderation filters and jailbreak detectors, are the primary safeguard against misuse in deployed AI applications, yet existing benchmarks are often vendor-biased, omit cost and latency, and rarely compare specialized guardrails against repurposed generalist LLMs.
Read next because BELLS-O: Evaluating the Operational Trade-offs of LLM Supervision Systems 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: phrase, under, eval, rate, compare, control. Source: arxiv cs.CR (Cryptography and Security).
arXiv:2606.20668v1 Announce Type: new Abstract: LLM supervision systems, namely input/output moderation filters and jailbreak detectors, are the primary safeguard against misuse in deployed AI applications, yet existing benchmarks are often vendor-biased, omit cost and latency, and rarely compare specialized guardrails against repurposed generalist LLMs. We present BELLS-O (Benchmark for the Evaluation of LLM Supervision Systems, Operational), the first independent operational benchmark of LLM supervision systems. BELLS-O evaluates 28 systems from 17 providers: every major specialized guardrail (e.g., LlamaGuard-4, ShieldGemma-2, Lakera Guard) and frontier generalists repurposed as supervisors (e.g., GPT-5.4, Claude Sonnet 4.6, Grok-4.1), jointly on detection rate, false-positive rate, latency, and monetary cost. We cover input/output moderation across 11 harm categories and jailbreak detection across 13 attack techniques, using in-house datasets built from handcrafted prompts, expert-curated samples, and quality-controlled synthetic generation. To suppress latent generator fingerprints in synthetic data, every generated sample is paraphrased. Mapping the Pareto frontier reveals use-case-dependent tradeoffs. On content moderation, specialized supervisors are operationally dominant: top systems match frontier LLMs on detection (~95% vs. 94%) at comparably low false-positive rates (<=2%), while running 5-10x faster and ~10x cheaper. On jailbreak detection, the tradeoff shifts: frontier LLMs achieve higher detection and lower false-positive rates but at 10-50x higher cost and 5-10x higher latency. We release the benchmark, framework, leaderboard, and datasets as the first vendor-neutral basis for selecting safeguards under real deployment constraints.
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 96arxiv cs.CL (NLP)arxiv:2606.20946unread
Scaling Diverse Language Generation for 3D Visual Grounding
Austin T. Wang, Dongchen Yang, Angel X. Chang · 2026-06-23
arXiv:2606. 20946v1 Announce Type: new Abstract: Developing robust models for 3D visual grounding (3DVG), the localization of entities in a 3D scene described in natural language, is important for enabling agents to correspond spatial language with objects in the physical world.
Read next because Scaling Diverse Language Generation for 3D Visual Grounding 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 "#351 follow-up: broader-vocab position-0 sweep at T=1.0 + position-1 suffix isolation", 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: rate, language, model. Source: arxiv cs.CL (NLP).
arXiv:2606.20946v1 Announce Type: new Abstract: Developing robust models for 3D visual grounding (3DVG), the localization of entities in a 3D scene described in natural language, is important for enabling agents to correspond spatial language with objects in the physical world. However, the lack of diverse descriptions at scale prevents models from generalizing beyond simple linguistic patterns. Recent such attempts lack diversity in the constraint types and language used to ground objects. Captioning methods cannot precisely contrast objects, which is important for visual grounding. We therefore propose ViGiL3D++, a scalable, scene-agnostic method that generates diverse visual grounding queries by combining constraint sampling in scene graphs with the language generation of LLMs. We show that it has greater diversity over existing scaled datasets and improves model performance over several 3DVG benchmarks but also illuminates outstanding limitations of VLMs.
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 limitation, limitations, benchmark.
- score 64M7 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.
Read next because Artifact verification caveats for Sagan clean results overlaps with experiment "Add C2 control arm (donor sees marker_B without marker_A) to disambiguate paired-marker binding from marker_B leaking alone". Matching terms: control. Source: M7 QA inline RSS threat source.
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.