[KongchangAI]
Benchmark

SWE Sweep

A class of software engineering benchmarks that provides issues from real code repositories, requiring models to locate defects, modify code, and pass tests, with scores based on the proportion of issues successfully fixed.

Source Articles