Paper

Steering Beyond the Support: Adversarial Training on Unsupervised Jailbroken Activation Simulation

Unreadunread

AI summary

arXiv:2605. 24535v1 Announce Type: new Abstract: Jailbreak prompts can trigger harmful completions on aligned LLMs, In accordance, safety steering has been proposed: test-time activation interventions that steer jailbreak activations to trigger refusal while preserving benign utility.

Reading status

Priority

About this paper

Authors: Luoyu Chen, Weiqi Wang, Zhiyi Tian, Chenhan Zhang, Feng Wu, Jianhuan Huang, Ahmed Asiri, Shui Yu
Released: 2026-05-26 (2mo ago)
Topic: Related to current project
Relation to my work: Read next because Steering Beyond the Support: Adversarial Training on Unsupervised Jailbroken Activation Simulation 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, latin, completions, rect, rate, test, completion. Source: arxiv cs.CR (Cryptography and Security).
arXiv: 2605.24535

Ask Claude about this paper

Posts a comment with @claude so the runner spins up a Claude Code session scoped to Steering Beyond the Support: Adversarial Training on Unsupervised Jailbroken Activation Simulation. Replies appear in the conversation below.

Comments

No comments yet.

Related records

No links yet.