# AtomicCommitBench: Can Coding Agents Reconstruct Commit Histories from Squashed Patches?

arXiv: https://arxiv.org/abs/2607.03332 · HTML: https://mlsanity.com/paper/2607.03332/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Software engineering coding agents, one of 16 directions in LLM agents and tool use.
- Builds on 20 of its 35 references that are in the map.
- First posted on arXiv 3 Jul 2026.

## Placed in

- LLM agents and tool use / [Software engineering coding agents](https://mlsanity.com/agents/software-engineering-coding-agents/)

## Builds on (most cited first, 20 of 20)

- [SWE-bench: Can Language Models Resolve Real-World GitHub Issues?](https://mlsanity.com/paper/2310.06770/) (2023)
- [SWE-agent: Agent-Computer Interfaces Enable Automated Software Engineering](https://mlsanity.com/paper/2405.15793/) (2024)
- [AutoCodeRover: Autonomous Program Improvement](https://mlsanity.com/paper/2404.05427/) (2024)
- [Multi-SWE-bench: A Multilingual Benchmark for Issue Resolving](https://mlsanity.com/paper/2504.02605/) (2025)
- [SWE-EVO: Benchmarking Coding Agents in Long-Horizon Software Evolution Scenarios](https://mlsanity.com/paper/2512.18470/) (2025)
- [Code Researcher: Deep Research Agent for Large Systems Code and Commit History](https://mlsanity.com/paper/2506.11060/) (2025)
- [From Commit Message Generation to History-Aware Commit Message Completion](https://mlsanity.com/paper/2308.07655/) (2023)
- [A Fine-grained Data Set and Analysis of Tangling in Bug Fixing Commits](https://mlsanity.com/paper/2011.06244/) (2020)
- [CommitBench: A Benchmark for Commit Message Generation](https://mlsanity.com/paper/2403.05188/) (2024)
- [RAG-Enhanced Commit Message Generation](https://mlsanity.com/paper/2406.05514/) (2024)
- [GitGoodBench: A Novel Benchmark For Evaluating Agentic Performance On Git](https://mlsanity.com/paper/2505.22583/) (2025)
- [AgentPack: A Dataset of Code Changes, Co-Authored by Agents and Humans](https://mlsanity.com/paper/2509.21891/) (2025)
- [HAFixAgent: History-Aware Program Repair Agent](https://mlsanity.com/paper/2511.01047/) (2025)
- [Atomizer: An LLM-based Collaborative Multi-Agent Framework for Intent-Driven Commit Untangling](https://mlsanity.com/paper/2601.01233/) (2026)
- [Code Change Characteristics and Description Alignment: A Comparative Study of Agentic versus Human Pull Requests](https://mlsanity.com/paper/2601.17627/) (2026)
- [LLM-Driven Collaborative Model for Untangling Commits via Explicit and Implicit Dependency Reasoning](https://mlsanity.com/paper/2507.16395/) (2025)
- [The effects of change decomposition on code review -- a controlled experiment](https://mlsanity.com/paper/1805.10978/) (2018)
- [Detecting Multiple Semantic Concerns in Tangled Code Commits](https://mlsanity.com/paper/2601.21298/) (2026)
- [EditFlow: Benchmarking and Optimizing Code Edit Recommendation Systems via Reconstruction of Developer Flows](https://mlsanity.com/paper/2602.21697/) (2026)
- [Lore: Repurposing Git Commit Messages as a Structured Knowledge Protocol for AI Coding Agents](https://mlsanity.com/paper/2603.15566/) (2026)

## Built on by (most cited first, 2 of 2)

- [Between the Commits: Process, Error, and Claim Reliability in a Wholly AI-Authored Codebase](https://mlsanity.com/paper/2609.29744/) (2026)
- [Runnable Commit Untangling for Coding Agents](https://mlsanity.com/paper/2610.11593/) (2026)

## Abstract (arXiv)

Coding agents often finish a session by returning one squashed patch that mixes feature implementation, bug fixes, refactorings, tests, and configuration edits. While the final code may be correct, collapsing unrelated edits into one patch removes the history structure needed for review, selective revert, and later maintenance. We study retrospective commit-history reconstruction: given a completed squashed change, an agent groups its hunks into commits and materializes a replayable commit sequence. We formalize the task as hunk-to-commit partitioning with a replay requirement and build AtomicCommitBench, containing 800 real consecutive-commit episodes from 10 Python projects. Because multiple decompositions may be reasonable, we evaluate outputs using complementary metrics: PPAR for replay validity, ARI for reference-based grouping quality, and TCR for failure containment on scoreable modified-test episodes. Natural retrospective reconstruction proves substantially harder than replay checking or synthetic tangling. Although nearly all non-random methods achieve replay validity (PPAR >= 0.988), grouping quality ranges from 0.03 to 0.46 ARI. Matched synthetic composites are much easier than real same-author squashed diffs (+0.333 ARI). In our evaluation, the GPT-5.4 setup driven by Codex CLI (0.46 ARI) and the GLM-5 setup driven by Claude Code (0.43 ARI) outperform MiniMax (0.31) and Kimi (0.29). Qualitative analysis identifies same-file lumping and support-hunk drift as recurring failure modes. Dependency-Aware Commit Evidence (DACE) improves the lower-scoring setups by 0.05 to 0.08 ARI, indicating that dependency cues and hunk-role information help agents avoid locality-driven grouping errors. AtomicCommitBench enables evaluation of the commit histories produced by coding agents alongside the final code.