# Harness-R1：9B模型修复智能体胜过397B

- 来源：Rohan Paul (@rohanpaul_ai)
- 发布时间：2026-08-05 12:07
- AIHOT 分数：32
- AIHOT 链接：https://aihot.virxact.com/items/cmsflgihx04i8rochmfatxml2
- 原文链接：https://x.com/rohanpaul_ai/status/2084853824922219007

## AI 摘要

Harness-R1 用 9B 模型重写冻结目标智能体的执行层，在修复任务上超越 397B 模型。该 9B 模型经监督微调和在线 RL 训练，依据可执行补丁是否提升目标智能体任务奖励来更新。目标模型不漂移，编辑器通过 GRPO 学习哪些补丁真正改善执行。

## 正文

A 9B model just outperformed a 397B one at repairing agents， and the reason is not size.

The 9B model had been specifically trained with supervised fine-tuning and online RL， receiving feedback from whether its executable patches actually improved the frozen target agent.

Harness-R1 trains one agent to rewrite the execution layer around another frozen agent.

Because a production agent's failed runs may already contain the data needed to improve its runtime.

Harness-R1 uses those failures to train a separate 9B harness engineer that edits the executable layer around a frozen target agent.

From each failure batch， it writes validated code hooks for episode setup， pre-decision guidance， pre-action mediation， or post-feedback recovery.

The patched target reruns the same tasks， and the measured change in task reward updates only the engineer through online GRPO.

So the target model never drifts under the reward signal， while the editor learns which patches improve execution rather than which ones merely look reasonable.

- arxiv. org/abs/2608.02276

Title： "Harness-R1： Learning to Edit Executable Runtime Harnesses from Agent Failure Trajectories"