SWE-Protégé: Learning to Selectively Collaborate With an Expert Unlocks Small Language Models as Software Engineering Agents
SWE-Protégé trains small language models to selectively seek expert guidance and avoid looping, achieving 42.4% Pass@1 on SWE-bench Verified with minimal expert use.
Published 2026Atlanta Poster Session 1 · Wed, Dec 9, 10:00 AM–1:00 PM local time · Hall C1▲ 2 on Hugging FacearXiv ↗OpenReview ↗
Only vote on papers you've read. Sign in with GitHub to vote.
SWE-Protégé delivers a striking 42.4% SWE-bench result by teaching a 7B model selective expert collaboration, though its sparse-assistance claim lacks ablation against self-consistency or failure analysis under wrong expert guidance.
Abstract
Small language models (SLMs) offer compelling advantages in cost, latency, and adaptability, but have so far lagged behind larger models on long-horizon software engineering tasks such as SWE-bench, where they suffer from pervasive action looping and low resolution rates. We introduce SWE-Protégé, a post-training framework that reframes software repair as an expert-protégé collaboration problem. In SWE-Protégé, an SLM remains the sole decision-maker while learning to selectively seek guidance from a strong expert model, recognize stalled states, and follow through on expert feedback. Our approach combines supervised fine-tuning on expert-augmented trajectories with agentic reinforcement learning that explicitly discourages degenerative looping and unproductive expert collaboration. We lightly post-train Qwen2.5-Coder-7B-Instruct to achieve 42.4% Pass@1 on SWE-bench Verified, a +25.4% improvement over the prior SLM state of the art, while using expert assistance sparsely (~4 calls per task and 11% of total tokens).