# SWE-PRBench: Benchmarking AI Code Review Quality Against Pull Request Feedback

arXiv: https://arxiv.org/abs/2603.26130 · HTML: https://mlsanity.com/paper/2603.26130/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 11 of its 22 references that are in the map.
- First posted on arXiv 27 Mar 2026.

## Builds on (most cited first, 11 of 11)

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Lost in the Middle: How Language Models Use Long Contexts](https://mlsanity.com/paper/2307.03172/) (2023)
- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment](https://mlsanity.com/paper/2303.16634/) (2023)
- [JudgeBench: A Benchmark for Evaluating LLM-based Judges](https://mlsanity.com/paper/2410.12784/) (2024)
- [Automating Code Review Activities by Large-Scale Pre-training](https://mlsanity.com/paper/2203.09095/) (2022)
- [Using Pre-Trained Models to Boost Code Review Automation](https://mlsanity.com/paper/2201.06850/) (2022)
- [Code Review Automation: Strengths and Weaknesses of the State of the Art](https://mlsanity.com/paper/2401.05136/) (2024)
- [DeepCRCEval: Revisiting the Evaluation of Code Review Comment Generation](https://mlsanity.com/paper/2412.18291/) (2024)
- [A ground-truth dataset and classification model for detecting bots in GitHub issue and PR comments](https://mlsanity.com/paper/2010.03303/) (2020)
- [LAURA: Enhancing Code Review Generation with Context-Enriched Retrieval-Augmented LLM](https://mlsanity.com/paper/2512.01356/) (2025)

## Built on by (most cited first, 2 of 2)

- [When AI Reviews Its Own Code: Recursive Self-Training Collapse in Code LLMs](https://mlsanity.com/paper/2606.28438/) (2026)
- [Adversarial Review: Structured Disagreement for Grounded Agentic Code Review](https://mlsanity.com/paper/2608.18167/) (2026)

## Abstract (arXiv)

We introduce SWE-PRBench, a benchmark of 350 pull requests with human-annotated ground truth for evaluating AI code review quality. Evaluated against an LLM-as-judge framework validated at kappa=0.75, 8 frontier models detect only 15-31% of human-flagged issues on the diff-only configuration, demonstrating that AI code review remains far below human expert performance despite strong results on code generation benchmarks. Pull requests are drawn from active open-source repositories, filtered from 700 candidates using a Repository Quality Score, and evaluated under three frozen context configurations: diff only (config_A), diff with file content (config_B), and full context (config_C), enabling systematic ablation of context provision strategies. All 8 models degrade monotonically from config_A to config_C, even when context is provided via structured semantic layers including AST-extracted function context and import graph resolution. The dominant mechanism is a collapse of Type2_Contextual issue detection at config_B, consistent with attention dilution in long contexts: a structured 2,000-token diff-with-summary prompt outperforms a 2,500-token full-context prompt enriched with execution context, behaviour mapping, and test signatures across all 8 models. The top four models are statistically indistinguishable (mean score 0.147-0.153) while a clear tier gap separates them from the remaining four (mean score <= 0.113). Dataset, contexts, annotations, and evaluation harness are released publicly.