Reproduce “Beyond Surface Metrics: Multi-Condition Requirements in Enterprise Maintenance and AI Coding Benchmarks” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Heeyong Eun
The authors compare task requirements across software-maintenance benchmarks using a taxonomy of fulfillment conditions, measuring condition breadth, pairwise co-occurrence, and source-grounded obligations. Among 1,177 tasks matched on issue classification, changed-file count, and code churn, mean Jaccard distance was 0.8138, and 57.4% shared no conditions. Matching controlled surface measures, not problem-solving difficulty.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.