SWE-Review: Closing the Loop on Issue Resolution with Agentic Code Review

2026-07-07

AI Summary

This work addresses the lack of systematic review and iterative refinement in pull requests (PRs) generated by current AI coding agents, which results in open-loop problem resolution. The paper proposes the first closed-loop PR review framework tailored to real-world software issues, leveraging an agent-driven approach that integrates codebase-aware context understanding, structured feedback generation, test-time scaling, and transfer learning to enable a “generate–review–revise” cycle. Key contributions include the SWE-Review-Bench evaluation benchmark and the SWE-Review-Traj dataset, which fill a critical gap in training data for review agents. Experimental results demonstrate that the proposed method significantly improves PR decision accuracy and post-revision issue resolution rates over single-round, fixed-context review baselines, while also enhancing the performance of downstream problem-solving models.

Abstract

Coding agents increasingly generate pull requests (PRs) for real-world software issues, yet one-shot PR generation remains open-loop: the PR is proposed without systematic review, diagnosis, or revision. We introduce \textbf{SWE-Review}, a framework for closing this loop with agentic code review. Given an issue and an AI-generated PR, a reviewer agent explores the repository, decides whether the PR should be accepted, and provides structured feedback for revision. We evaluate this setting with our proposed \textbf{SWE-Review-Bench} to measure both review correctness and downstream revision usefulness. We further curate \textbf{SWE-Review-Traj} dataset to study broader applications of agentic review and fill the data-scarcity gap for open reviewer training. Experiments show that agentic review continuously improves PRs through a generate-review-revise loop, outperforms single-turn fixed-context review in both decision accuracy and resolve rate after revision, transfers beyond review to improve issue-resolution models, and enables effective and efficient test-time scaling. These results position agentic code review as a practical mechanism for moving AI coding agents from one-shot PR generation toward closed-loop issue resolution.

Problem

• code review
• pull request
• issue resolution
• AI coding agents
• closed-loop

Innovation

• agentic code review
• closed-loop issue resolution
• SWE-Review
• code revision
• AI coding agents