Safe Evolution with Circuit Anchors

If you run self-improving model loops, capability gains can quietly erode safety behavior; constraining a small mechanistically identified circuit preserved it more cheaply than reward-based penalties.

• articleOne Recipe, Many Harnesses: What Self-Evolution Encodes Across Languages and ModelsSiqi Yang, Qianlan Yang, Yu-Xiong Wang, Saurabh Pujar, Martin Hirzel
• articleSelf-Evolving Embodied Agents via Skill-Harness EvolutionPeidong Wang, Zhiming Ma, Ying Chang, Xufang Luo, Xiaocui Yang, Shi Feng, Yuqing Yang, Dongsheng Li
• articleWhere Does Harness-Optimization Value Live? Localized Gains and the Budget-Splitting Trap in Self-Evolving LLM AgentsMichael Nguyen, Wei Chen Tan, Nurul Aisyah Hassan, Arvind Raman, Li Hua Lim, Ahmad Faiz Razak

Checking sign-in…

Loading comments…