On Safety Risks in Experience-Driven Self-Evolving Agents

arXiv:2604.16968v1 Announce Type: new Abstract: Experience-driven self-evolution has emerged as a promising paradigm for improving the autonomy of large language model agents, yet its reliance on self-curated experience introduces underexplored safety risks. In this study, we investigate how experience accumulation and utilization in self-evolving agents affect safety performance across web-based and embodied environments. Notably, experience gathered solely from benign tasks can still compromise safety in high-risk scenarios. Further analysis attributes this degradation to the execution-oriented nature of accumulated experience, which reinforces agents' tendency to act rather than refuse. In more realistic settings where agents encounter both benign and harmful tasks, refusal-related experience mitigates safety decline but induces over-refusal, revealing a fundamental safety-utility trade-off. Overall, our findings expose inherent limitations of current self-evolving agents and call for more principled strategies to ensure safe and reliable adaptation.

Related

• Why Agents Compromise Safety Under Pressure
• When Choices Become Risks: Safety Failures of Large Language Models under Multiple-Choice Constraints
• AAPO: Enhancing the Reasoning Capabilities of LLMs with Advantage Margin
• Large Language Model Post-Training: A Unified View of Off-Policy and On-Policy Learning
• Reverse Constitutional AI: A Framework for Controllable Toxic Data Generation via Probability-Clamped RLAIF

Source: arXiv cs.CL | 2026-04-21