Harness-S: Learning from Sibling Harnesses for Generalizable Agent Training
Paper · https://arxiv.org/abs/2611.04277
Harness evolution from a common base exhibits Locality, Complementarity, and Nonstationarity: variants repair different failures, and their utility changes as the policy learns. Rather than seeking a universally superior harness, we introduce Harness-S to selectively rescue failures of a base execution system through sibling harnesses and transfer the resulting capabilities into a single policy that remains effective under the original harness. Directional Harness Evolution develops specialized sibling harnesses along distinct repair directions, forming a reusable Sibling Harness Pool that remains fixed during policy training. Complementary Harness Rescue Training invokes this pool only when all initial attempts under the base harness fail. The same policy generates fresh rollouts under selected auxiliary harnesses, and these trajectories replace the failed group for reinforcement learning. An Adaptive Harness Router selects the Top-k variants using Harness Behavior Profiles and voting across multiple LLMs. Policy-aware profile re-anchoring refreshes this evidence as the policy learns. Across AppWorld, SWE-Gym, and DBBench, the trained policy substantially outperforms policy-only training when returned to the original base harness. Behavioral analyses and held-out cases support the transfer of harness-elicited behaviors into the policy itself. Reusing the router for Rank-1 inference provides additional gains at the same executor rollout count.
First seen by Oriel on .
AI summary
Harness-S trains a single policy by selectively using sibling execution harnesses to rescue failures, transferring their capabilities while preserving performance under the original harness.
- Directional Harness Evolution creates specialized sibling harnesses for distinct repair directions and keeps the resulting Sibling Harness Pool fixed during training.
- Complementary Harness Rescue Training invokes auxiliary harnesses only after all initial base-harness attempts fail, replacing failed rollout groups with successful trajectories for reinforcement learning.
- An Adaptive Harness Router selects Top-k variants using Harness Behavior Profiles and multi-LLM voting, while policy-aware re-anchoring refreshes profiles during learning.
- Across AppWorld, SWE-Gym, and DBBench, Harness-S substantially outperforms policy-only training under the original harness and improves Rank-1 inference at equal executor rollout counts.
#agent training #reinforcement learning #llm agents #harness evolution
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Points
Score reported by the source.
- Hugging Face Papers: 67, latest observation ; +15 over observed period since
Comments
Comment count reported by the source.
- Hugging Face Papers: 5, latest observation ; +2 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Papers | 2611.04277 | https://arxiv.org/abs/2611.04277 Read discussion |