Jingxuan Xu is the first listed author of SWE-Compass, a benchmark that evaluates coding agents on tasks drawn from GitHub pull requests.[1][2]
Start with SWE-Compass →Links
Hugging FaceContributions
Tools, writing & talks
Writing, research & talks
Start here
SWE-Compassarxiv.orgA benchmark for coding agents built from 2,000 software engineering tasks.
Start with the benchmark design to see which engineering tasks it measures.
First listed author[1]