HuggingFace Daily Papers·· 2 天前
HarnessAdapt-RL：训练编码智能体时同步演化 harness
Training the Whole Agent, Not Just the Model
AI 导读
HarnessAdapt-RL 在强化学习中同时训练策略并自适应调整 harness；对修复策略缺陷由冻结教师生成任务级机制技能，对跨任务交互问题由另一教师模型演化工具与 hooks。
正文 · 原文
Abstract:Training coding agents typically updates the policy while keeping the harness fixed. During training, trajectories reveal flawed repair strategies and interaction problems that vary across tasks and change as the policy learns. We introduce HarnessAdapt-RL, which trains the policy and adapts its harness during reinforcement learning. For flawed repair strategies, a frozen teacher generates task-specific mechanism-level skills. For interaction problems recurring across tasks, a separate teacher model evolves tools and hooks, keeping each change only if it improves the observed pass rate or the observed trainable rate without lowering either. The policy trains with the adapted harness, and its new trajectories guide further adaptation. With a 9B base model, HarnessAdapt-RL improves pass@1 over GRPO by 9.8, 12.6, and 6.4 percentage points on SWE-bench Verified, Multilingual, and Pro, and outperforms GRPO with the human-designed OpenHands harness. Gains from mechanism-level skills persist after their removal at evaluation. Re-evolving tools and hooks as the policy changes outperforms freezing early versions. Removing them at deployment, in contrast, costs up to 7.7 percentage points. Training therefore produces a model-harness pair: the trained policy and the evolved tools and hooks, with skills removed at deployment.
| Subjects: | Software Engineering (cs.SE) |
| Cite as: | arXiv:2610.28069 [cs.SE] |
| (or arXiv:2610.28069v1 [cs.SE] for this version) | |
| https://doi.org/10.48550/arXiv.2610.28069 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Lingxiao Tang [view email]
[v1]
Tue, 20 Oct 2026 11:51:23 UTC (486 KB)
来源：HuggingFace Daily Papers · arxiv.org