Co-Evolving Harnesses and Models: On-Policy Correction Helps Weaker Models Catch Up Where Imitation Fails
Research shows on-policy expert correction, not imitation fine-tuning, lets weaker agent models catch up under evolved harnesses.
Researchers study how to combine automated agent-harness evolution with lightweight fine-tuning across seven enterprise agent tasks. Naively training weaker models (Qwen3-Coder, Gemma 4) on expert trajectories under an evolved harness regressed performance by 4 to 30 points on all tasks. They propose an on-policy correction pipeline, automated by a meta-level MLE agent, where an expert rewrites only the failing turn of the weaker model's rollout, preserving model-harness fit.
- Naive imitation under an evolved harness regressed performance 4-30 points across all seven tasks
- Imitation disrupts model-harness fit: weaker models adopt expert planning without competence to execute it
- On-policy correction rewrites only failing turns, combining harness evolution and weight adaptation
- Smaller models with evolved harnesses target frontier-model agentic performance at lower cost
- AI models
- Qwen3-CoderGemma 4
From the original article
1 min read at arxiv.orgAgent harnesses (the system prompt, tool set, execution hooks, and context-management scaffolding around a model) are a critical determinant of agentic task success. Automated harness evolution can enable smaller models to perform well on domain-specific tasks at a fraction of frontier-model cost.
By Zhou Yu · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.