DANGERLENS is introduced, a probe that maps hidden states into vocabulary space and uses blinded semantic judgments to distinguish harmful goals and operational means from hazard recognition and safety control and introduces a direct audit of intermediate representations.
Tian-Zhuo Yang, Shao-Heng Yan, Lin Lu et al.· 0 citations
Foundation models supply reinforcement learning (RL) with priors that mitigate its longstanding weaknesses in sample efficiency and transfer, but their token-by-token generation makes queries sequential and costly. Jev, a recently released decision model, generates nothing and returns calibrated, typed answers in a sin...
Yi Ma, Tian-Pei Yang, Yao-Dong Yang et al.· 1 citation
This work introduces AgentBound, the first four-way counterfactual generation-and-evaluation framework for tool-using agent safety, and shows that effective agentic alignment requires action decisions to track permission-relevant execution evidence, rather than refusal strength alone.