Skip to content

SafeCoEvo: Co-Evolving Safety Harnesses and Guards for LLM Agents at Test-Time

Sep 2026 · arXiv.org · Vol abs/2609.36580 · 3 citations · 61 references
Computer Science

TL;DR

SafeCoEvo is proposed, a test-time Harness-Guard co-evolution framework for LLM agent safety that enables the external safety system to continually adapt from accumulated runtime experience to achieve simultaneous gains in safety and task utility.

Abstract

LLM agents deployed in real-world environments continually encounter new tasks and safety risks, while execution feedback typically becomes available only after each task is completed. However, existing self-evolving approaches commonly rely on multiple rounds of optimization over fixed and repeatedly accessible task distributions, fundamentally differing from test-time adaptation in real-world deployment, where only experience accumulated from past tasks can be used to improve safety decisions on future unseen tasks. To address this limitation, we propose SafeCoEvo, a test-time Harness-Guard co-evolution framework for LLM agent safety that enables the external safety system to continually adapt from accumulated runtime experience. SafeCoEvo jointly improves two complementary safety capabilities at different timescales: S-Harness rapidly externalizes recent runtime experience into updatable explicit safety knowledge that can promptly influence subsequent tasks, while GuardVPO internalizes accumulated runtime safety experience over a longer timescale into parametric risk-judgment capabilities. By combining short-term rapid adaptation with long-term capability consolidation, SafeCoEvo continually improves the agent's safety capabilities, reducing the unsafe outcome rate by 10.05% while improving the task success rate by 12.15% over the strongest baseline, thereby achieving simultaneous gains in safety and task utility.

View source

Similar papers

#artificial intelligence Preprint Sep 2026

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

Through harness-policy co-evolution, SafeEvolve converts safety experience into an evolved runtime harness and improved policy behavior, and experiments show that SafeEvolve achieves a stronger safety-utility tradeoff than existing baselines.

Qing-Hua Mao, Wanying Qu, Da-Di Guo et al. · 7 citations · ⚡2
#artificial intelligence Review Sep 2026

Safety in Self-Evolving Agents: A Survey

SAVER is introduced, a transition-centered framework in which Substrate locates reusable influence, Adaptation captures how it changes, Violation identifies compromised safety attributes, Exposure marks where failures become observable, and Response assesses containment, repair, or revocation.

Jia-Hao Chen, Zhou Feng, Ou-Bo Ma et al. · 1 citation
#artificial intelligence Preprint Oct 2026

VEPA: Verified Evolving Formal Policy for Lifelong LLM Agent Guardrails

VEPA is introduced, a framework for the verified evolution of formal agent guardrails that represents the guardrail as a formal policy of executable symbolic rules over typed agent events and consistently outperforms baselines while expanding the scope of deterministic formal-policy enforcement.

Duy Nguyen, Lesly Miculicich, Bhavana Dalvi et al. · 0 citations
#artificial intelligence Preprint Aug 2026

openJiuwen: Beyond Static Harnesses for Long-Horizon Coding Agents

OpenJiuwen provides a shared execution substrate and Rail-based capability composition across single agents, delegated sub-agents, and Swarm Flow, enabling developers to construct sophisticated agent harnesses under common execution semantics.

openJiuwen Team Tao Yu, Xin-Yu Zhang, Qian-Qian Chen et al. · 0 citations
#natural language process... Preprint Sep 2026

EvoHarnessBench: Can Your Agents Keep Pace with an Evolving Harness?

EvoHarnessBench is introduced, a benchmark for evaluating agents under controlled harness evolution across three axes (tools, skills, and agents), and establishes harness evolution as a distinct challenge for building agents that can keep pace with an evolving harness while preserving previously effective behavior.

Zi-Xuan Ke, Vaidehi Patil, Hai-Zhou Shi et al. · 5 citations · ⚡1

Related blog posts

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.