# CMU paper trains a 4B proposer to edit agent harnesses without changing the solver

The harness-learning paper uses reinforcement learning so a proposer revises harness code from a task, the current harness, and an execution report. A trained 4B proposer beats its 35B teacher on single-step Reasoning Gym revisions, and HotpotQA training transfers to MuSiQue and 2WikiMultihopQA.

Language: en
Time zone: America/Los_Angeles
Current time (UTC, cached as of): 2026-10-23T06:45:00.000Z
Query date (America/Los_Angeles): 2026-10-22

HTML: https://didcodexreset.com/news/8765e117d8bf4962af57d0d5.html

Content language: en
Localization state: sameLanguage

Source: elvis · Published 10/3/2026, 17:20:53

A CMU paper, [Harness Learning Enables Generalizable Test-Time Adaptation](https://arxiv.org/abs/2609.35738), trains a proposer model with reinforcement learning to improve an agent by editing its harness code rather than changing the solver model’s weights.

The proposer reads a task, the current harness, and an execution report, then writes a code edit. Its reward is the score of the revised harness, and the solver model stays fixed.

A trained 4B proposer beats its 35B teacher at single-step revision on Reasoning Gym, including task families it never saw in training. A proposer trained on HotpotQA keeps improving harnesses on MuSiQue and 2WikiMultihopQA.

Tags: Harness learning, CMU, AI agents, Reinforcement learning, Reasoning Gym

[View original post](https://x.com/omarsar0/status/2106529236068819394)