Metraipolis › archive › dc_8730af5f46ea
SafeEvolve: Harness-Policy Co-Evolution from Agent Experience for Safety Alignment — arXiv:2609.02786
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.02786
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T03:26:44.521Z
- SHA-256 of the archived text
5fa524d1a438e38c3cba20046b5e87415dc528b25c5b7cffe7345d511f503b88- Stable URL of this page
https://metraipolis.com/documento/dc_8730af5f46ea
How to cite
SafeEvolve: Harness-Policy Co-Evolution from Agent Experience for Safety Alignment — arXiv:2609.02786 — arXiv (en), https://arxiv.org/abs/2609.02786 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_8730af5f46ea (Metraipolis archive dc_8730af5f46ea, sha256 5fa524d1a438e38c3cba20046b5e87415dc528b25c5b7cffe7345d511f503b88)
BibTeX
@misc{dc8730af5f46ea,
title = {SafeEvolve: Harness-Policy Co-Evolution from Agent Experience for Safety Alignment — arXiv:2609.02786},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.02786},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_8730af5f46ea; sha256 5fa524d1a438e38c3cba20046b5e87415dc528b25c5b7cffe7345d511f503b88},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.02786. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
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 framework for agent safety alignment. SafeEvolve leverages safety experience from completed on-policy trajectories to drive a continual loop of harness-policy co-evolution. On the harness side, SafeEvolve converts trajectory-level safety evidence into bounded, component-level updates across safety prompt and hierarchical skills, yielding auditable and reversible harness artifacts. On the policy side, SafeEvolve follows a two-stage SFT-RL paradigm, where harness-use SFT bootstraps the policy to actively leverage evolved harness artifacts, and harness-augmented RL further shapes autonomous safety behaviors during multi-step exploration via verifier-decomposed rewards. Through harness-policy co-evolution, SafeEvolve converts safety experience into an evolved runtime harness and improved policy behavior. Experiments on agentic safety benchmarks show that SafeEvolve achieves a stronger safety-utility tradeoff than existing baselines. For Qwen3.5-4B, SafeEvolve achieves a $3\times$ ASR reduction on AgentDojo while improving benign utility from 59.79% to 61.86%. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-03 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI) ; Cryptography and Security (cs.CR). Pagina del lavoro: https://arxiv.org/abs/2609.02786 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-17T03:13:12.453Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-03?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).