# Harness-Zero 论文提出 agent-as-harness 方法，将专用 harness 能力蒸馏进模型

- 来源：elvis (@omarsar0)
- 发布时间：2026-09-24 20:13
- AIHOT 分数：52
- AIHOT 链接：https://aihot.news/items/cmufisajc03opro6owzj4r56e
- 原文链接：https://x.com/omarsar0/status/2103095360239636666

## AI 摘要

北京大学、Google 和 HKUST 等机构的论文 Harness-Zero 研究 agent harness 蒸馏，通过 agent-as-harness 在训练时用优化后的 harness 纠正学生模型在目标动作空间的响应，生成训练示范后微调。

## 正文

Super interesting paper from Google and colleagues.

It studies where it's possible to distill an agent harness.

With the specialized harness removed, macro task success goes from 23.3% to 44.3%. That is higher than the 41.7% the base model reaches with the harness attached.

Harness-Zero uses the optimized harness only during training.

The optimized harness and the deployment harness have different action spaces, so a harnessing agent guided by the optimized harness corrects the student's responses in the deployment action space before they run. Those corrected runs become the training demonstrations.

Across 28 harness-induced behaviors in knowledge work, tool use and science, 82.3% are recovered on average. For frontier models using the same evolved harness, the agent-as-harness form also beats the code-as-harness form.

It remains to be seen how robust the approach is, but it's very interesting to see potential in harness distillation.

Paper: https://arxiv.org/abs/2609.24974

Chat with Paper: https://academy.dair.ai/papers/harness-zero-harness-distillation-via-agent-as-harness-2609.24974