Harness Learning Lets Agents Adapt at Test Time Without Parameter Updates

A paper from MIT, Meta, and Yale proposes ReHarness and Harness Learning, enabling LLM agents to adapt and learn continuously at test time without parameter updates by revising their harness through backward experience flows.

2026-10-12 ~ 2026-10-12 · 2 related posts

• Harness Learning: RL-Trained Proposer Adapts Agent Harnesses at Test Time Without Weight Updates — pmddomingos · 2026-10-12
• Meta's ReHarness adds backward revision to agent harnesses, beating memory and reflection baselines — Yuxin Xiao · 2026-10-12