{"type":"analysis","id":"analysis:research-fine-tuning-a-small-model-on-expert-trajectories-can-undo-the-gains-from-an","slug":"research-fine-tuning-a-small-model-on-expert-trajectories-can-undo-the-gains-from-an","title":"Fine-tuning a small model on expert trajectories can undo the gains from an evolved harness. Correcting single turns in its own rollouts avo","description":"A Salesforce AI preprint finds that LoRA imitation of Gemini 3.1 Pro under a harness evolved for Qwen3-Coder cut success by 4–30 points on all seven tasks. Correcting single turns preserved the harness gains and added only about 1.7 points on average. The negative result is the part builders should act on.","published_at":"2026-10-22T23:00:00.000Z","updated_at":"2026-10-22T23:00:00.000Z","record_date":"2026-10-22","topics":["gemini","tools","research"],"source_urls":["https://arxiv.org/abs/2607.08938","https://arxiv.org/html/2609.09134","https://arxiv.org/abs/2609.09134","https://arxiv.org/abs/2607.22688","https://arxiv.org/abs/2609.24974","https://arxiv.org/abs/2604.25850"],"language":"en","routes":{"html":"https://newruntime.com/analysis/research-fine-tuning-a-small-model-on-expert-trajectories-can-undo-the-gains-from-an/","markdown":"https://newruntime.com/analysis/research-fine-tuning-a-small-model-on-expert-trajectories-can-undo-the-gains-from-an.md","json":"https://newruntime.com/analysis/research-fine-tuning-a-small-model-on-expert-trajectories-can-undo-the-gains-from-an.json"},"author":{"role":"author","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"editorial_provenance":{"desk":"authorial","target_ref":"deep_research:4db8a6a5-4654-4efa-b9b5-a1ca61605a25","authorial_article_id":"4db8a6a5-4654-4efa-b9b5-a1ca61605a25"},"next_reads":[{"type":"topic","path":"/topics/gemini/","url":"https://newruntime.com/topics/gemini/","title":"Gemini","reason":"Explore the gemini topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/gemini.md","json_url":"https://newruntime.com/topics/gemini.json"},{"type":"topic","path":"/topics/tools/","url":"https://newruntime.com/topics/tools/","title":"Tools","reason":"Explore the tools topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/tools.md","json_url":"https://newruntime.com/topics/tools.json"},{"type":"topic","path":"/topics/research/","url":"https://newruntime.com/topics/research/","title":"Research","reason":"Explore the research topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/research.md","json_url":"https://newruntime.com/topics/research.json"},{"type":"collection","path":"/analysis/","url":"https://newruntime.com/analysis/","title":"Analysis","reason":"Open the latest authorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/analysis.json"},{"type":"collection","path":"/source-ledger.json","url":"https://newruntime.com/source-ledger.json","title":"Source Ledger","reason":"Inspect the public source evidence graph.","media_type":"application/json"}]}