--- title: "Step Rejection Fine-Tuning: Squeezing More Signal from Noisy Agent Trajectories" url: https://daily.dev/posts/step-rejection-fine-tuning-squeezing-more-signal-from-noisy-agent-trajectories-884rq1xoj source_url: https://blog.jetbrains.com/research/2026/06/step-rejection-fine-tuning type: article source: "JetBrains" published: 2026-06-17T09:23:07.706Z updated: 2026-06-17T10:21:09.166Z tags: ["ai-agents", "reinforcement-learning"] reading_time: 13 upvotes: 1 comments: 0 language: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # Step Rejection Fine-Tuning: Squeezing More Signal from Noisy Agent Trajectories **[JetBrains](https://daily.dev/sources/jetbrains)** · 13 min read · 1 upvotes · 0 comments ## Summary JetBrains Research introduces Step Rejection Fine-Tuning (SRFT), a technique to extract learning signal from failed LLM agent trajectories. Standard Rejection-sampling Fine-Tuning (RFT) discards entire runs that don't produce a correct final answer, wasting up to 61% of training data. A manual analysis of failed runs found that only ~24% of steps in failed trajectories are actually harmful — the rest are productive. SRFT uses a cheap LLM critic to label each step in unsuccessful trajectories as good, unnecessary, mistake, or recover, then masks the loss on mistake steps during training. This lets the model learn useful behaviors from failed runs without reinforcing bad habits. Experiments on SWE-bench Verified with Qwen2.5-Coder-32B show SRFT achieves 32.2% resolution vs. 30.9% for standard RFT, and even training solely on masked unsuccessful trajectories (29.7%) outperforms naively mixing resolved and unresolved data (28.5%). The paper was accepted to the DL4C workshop at ICML 2025. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Similar posts on daily.dev - [OpenAI at QCon AI NYC: Fine Tuning the Enterprise](https://daily.dev/posts/openai-at-qcon-ai-nyc-fine-tuning-the-enterprise-gbqxbdmcn) · InfoQ · 1 upvotes · 0 comments - [Post-Training Generative Recommenders with Advantage-Weighted Supervised Finetuning](https://daily.dev/posts/post-training-generative-recommenders-with-advantage-weighted-supervised-finetuning-aaj5ljika) · Netflix TechBlog · 1 upvotes · 0 comments --- Tags: [#ai-agents](https://daily.dev/tags/ai-agents), [#reinforcement-learning](https://daily.dev/tags/reinforcement-learning) [View this post on daily.dev](https://daily.dev/posts/step-rejection-fine-tuning-squeezing-more-signal-from-noisy-agent-trajectories-884rq1xoj)