Type: url
Original: https://arxiv.org/pdf/2601.17581
Status: generated
Summary
“How AI Coding Agents Modify Code: A Large-Scale Study of GitHub Pull Requests” (Ogenrwot and Businge; MSR ‘26) compares 24,014 merged agent-authored pull requests with 5,081 merged human-authored ones from the MSR 2026 Mining Challenge version of the AIDev dataset. Agentic PRs differ most from human PRs in commit count (a large effect) and moderately in files touched and deleted lines, tending toward smaller, more localized edits, with variation between agents. Measured by lexical and semantic similarity, agentic PR descriptions align with their diffs at least as well as, and slightly better than, human descriptions.