Teaching Agents to Code Reliably

2026-10-02

AI Summary

This study addresses the inefficiency of autonomous coding agents in code repair, which stems from insufficient search diversity and unreliable verification signals. To overcome these limitations, this work proposes a "teachability" paradigm that leverages execution feedback to guide search space exploration and introduces reverse tree scoring to enhance verification. Furthermore, the verifier is jointly optimized through weighted supervised fine-tuning (SFT) and reinforcement learning (RL), effectively internalizing reliable repair behaviors into the policy model. Evaluated on the SWE-bench Verified benchmark, the proposed approach increases the pass rate to 52.8% while reducing the average number of repair steps by 52%. It achieves a pass@1 of 43.0% with robust performance across models of varying scales, significantly mitigating verification false positives.

Abstract

Autonomous coding agents solve repository issues by reading code, running commands, editing files, and submitting patches. Extra inference-time compute yields gains only when it produces a useful repair and supplies reliable evidence for choosing one. Three behaviors decide both, and we argue they are teachable rather than byproducts of scale, so a policy can carry them instead of a scaffold. Location diversity remains narrow, since attempts return to the same site and extra samples add no coverage. Edit diversity is left unexploited, since methodologies that differ resolve complementary issues no single run reaches. Verification misleads, since a test the agent writes for its own patch accepts many incorrect ones. Directing search by execution feedback and scoring each patch against its own reverted tree resolves 52.8% of SWE-bench Verified using 48.1% of the agent-steps an eight-sample baseline spends. Training moves these behaviors into the policy. On the 270 issues held out from SFT and RL training, weighted supervised fine-tuning raises pass@1 from 31.9% to 35.2% and pass@8 from 46.7% to 51.1%. A reinforcement objective then trains the verifier against gold-labeled repairs and incorrect variants, crediting the assertions that detect them. It raises pass@1 to 43.0% and pass@8 to 60.7%, lifts verifier precision from 26.8% to 41.7%, and more than halves false acceptance. Resolution improves on two of three out-of-distribution suites and verifier precision on all three, and the gains hold at 7B, 14B, and 30B against published coder baselines.

Problem

• autonomous coding agents
• inference-time compute
• location diversity
• edit diversity
• verification reliability

Innovation

• Autonomous coding agents
• Inference-time compute
• Reinforcement learning
• Patch verification
• SWE-bench