paperSeptember 2026Unreviewed

SafeEvolve: Harness-Policy Co-Evolution from Agent Experience for Safety Alignment

Qinghua Mao, Wanying Qu, Dadi Guo, Leitao Yuan, Qingyu Liu, Yu Li, Guanxu Chen, Yanwei Fu, Xi Lin, Xia Hu, Dongrui Liu

Abstract

The performance of LLM-based agents is jointly shaped by the base model and the harness used when interacting with the environment. This exposes them to safety risks in both harmful final responses and multi-step execution trajectories. Existing safety alignment mechanisms often rely on either external harness updates or policy optimization, yet applying either paradigm in isolation fails to bridge runtime control with intrinsic safety. We propose SafeEvolve, an experience-driven self-evolving f

Cite

@misc{mao2026safeevolve,
  title = {{SafeEvolve: Harness-Policy Co-Evolution from Agent Experience for Safety Alignment}},
  author = {Qinghua Mao and Wanying Qu and Dadi Guo and Leitao Yuan and Qingyu Liu and Yu Li and Guanxu Chen and Yanwei Fu and Xi Lin and Xia Hu and Dongrui Liu},
  year = {2026},
  month = sep,
  eprint = {2609.02786},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.02786}
}