Atomizer: An LLM-based Collaborative Multi-Agent Framework for Intent-Driven Commit Untangling

TL;DR

Atomizer is a multi-agent framework that uses large language models and iterative refinement to improve the untangling of composite commits by understanding semantic intent, outperforming existing graph-based methods.

Contribution

It introduces a novel collaborative multi-agent approach with intent inference and iterative refinement, addressing semantic understanding and single-pass limitations in commit untangling.

Findings

Outperforms state-of-the-art graph-based methods by over 6% on benchmark datasets.

Achieves over 16% improvement on complex commits.

Demonstrates significant performance gains on C# and Java datasets.

Abstract

Composite commits, which entangle multiple unrelated concerns, are prevalent in software development and significantly hinder program comprehension and maintenance. Existing automated untangling methods, particularly state-of-the-art graph clustering-based approaches, are fundamentally limited by two issues. (1) They over-rely on structural information, failing to grasp the crucial semantic intent behind changes, and (2) they operate as ``single-pass'' algorithms, lacking a mechanism for the critical reflection and refinement inherent in human review processes. To overcome these challenges, we introduce Atomizer, a novel collaborative multi-agent framework for composite commit untangling. To address the semantic deficit, Atomizer employs an Intent-Oriented Chain-of-Thought (IO-CoT) strategy, which prompts large language models (LLMs) to infer the intent of each code change according to both the structure and the semantic information of code. To overcome the limitations of ``single-pass'' grouping, we employ two agents to establish a grouper-reviewer collaborative refinement loop, which mirrors human review practices by iteratively refining groupings until all changes in a cluster share the same underlying semantic intent. Extensive experiments on two benchmark C# and Java datasets demonstrate that Atomizer significantly outperforms several representative baselines. On average, it surpasses the state-of-the-art graph-based methods by over 6.0% on the C# dataset and 5.5% on the Java dataset. This superiority is particularly pronounced on complex commits, where Atomizer's performance advantage widens to over 16%.