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

Hugging Face Daily Papers Papers
agentic-code-review issue-resolution code-quality ai-agents pull-request-revision test-time-scaling swe-bench

Summary

This paper introduces SWE-Review, a framework that closes the loop on AI-generated pull requests by enabling iterative agentic review and revision cycles, improving code quality and issue resolution. Experimental results show that it outperforms single-turn review and enables effective test-time scaling.

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 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 SWE-Review-Bench to measure both review correctness and downstream revision usefulness. We further curate 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.
Original Article
View Cached Full Text

Cached at: 07/09/26, 07:53 AM

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

Source: https://huggingface.co/papers/2607.06065

Abstract

Agentic code review framework enhances AI-generated pull requests through iterative review and revision cycles, improving both code quality and issue resolution capabilities.

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 SWE-Review, a framework for closing this loop withagentic 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 SWE-Review-Bench to measure bothreview correctnessand downstreamrevision usefulness. We further curate 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 agenerate-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 efficienttest-time scaling. These results positionagentic code reviewas a practical mechanism for moving AI coding agents from one-shot PR generation toward closed-loopissue resolution.

View arXiv pageView PDFProject pageGitHub8Add to collection

Get this paper in your agent:

hf papers read 2607\.06065

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper2

#### SWE-Lego/SWE-Review-8B Text Generation• 308k• Updatedabout 1 hour ago • 21 • 1 #### SWE-Lego/SWE-Review-30B-A3B Text Generation• 211k• Updatedabout 1 hour ago • 14

Datasets citing this paper2

#### SWE-Lego/SWE-Review-Traj Viewer• Updatedabout 1 hour ago • 8.91k • 208 • 1 #### SWE-Lego/SWE-Review-Bench Viewer• Updatedabout 1 hour ago • 1.38k • 63

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2607.06065 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Agentic Code Review (15 minute read)

TLDR AI

An analysis of how AI coding agents have shifted the bottleneck from writing code to reviewing it, with data showing a 861% increase in code churn and a rise in defect rates, making code review the most leveraged skill in software engineering.

Make Reviews Possible Again With This One Simple Trick

Lobsters Hottest

The article proposes using stacked branches (small, sequential pull requests) to make reviewing AI-generated code more manageable and effective, addressing the common problem of large, hard-to-review diffs.