The authors curated 19 software engineering tasks from real-world repositories and represented each task as an explicit dependency graph.
Coding skill does not guarantee effective multi-agent software collaboration
AsynCodeBench tracks whether and when coding agents satisfy explicit cross-component dependencies, separating coordination quality from final test performance.
Academic
Kaituo Zhang · Zhen Xiong · Zhimeng Jiang · Mingyu Zhong · Zhouyuan Yuan · Zhecheng Li · +5 more
University of Houston · New York University · Texas A&M University · UCSD · Worcester Polytechnic Institute
Research Digest··2 min read
The authors introduce a benchmark for evaluating collaboration among asynchronous coding agents using executable checks for dependencies between their assigned components.
Why this paper
From University of Houston and 5 others
In one line
Multi-agent collaboration in software engineering is distinct from coding ability and measurable via dependency resolution metrics ADPR and DRS.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§