SWE-Lego/SWE-Review-Bench
Viewer • Updated • 1.38k • 69
SWE-Review
SWE-Review: Closing the Loop on Issue Resolution with Agentic Code Review. Benchmark, trajectories, training data, and models for agentic code review.
Note 1,500 AI-generated PRs across 3 splits for evaluating code review agents
SWE-Lego/SWE-Review-Traj
Viewer • Updated • 8.91k • 114 • 1Note 8,914 decision-correct agentic review trajectories with judge quality scores
SWE-Lego/SWE-Review-8B
Text Generation • 308k • Updated • 18Note Qwen3-8B fine-tuned for agentic code review
SWE-Lego/SWE-Review-30B-A3B
Text Generation • 211k • Updated • 15Note Qwen3-30B-A3B (MoE) fine-tuned for agentic code review