Debug2Fix: Can Interactive Debugging Help Coding Agents Fix More Bugs?

TL;DR

Debug2Fix introduces an interactive debugging framework for coding agents, significantly enhancing bug fixing performance by integrating debuggers into the agent architecture.

Contribution

The paper presents a novel subagent architecture that incorporates debuggers into coding agents, improving bug fixing capabilities and performance on benchmark datasets.

Findings

Achieved over 20% performance improvement on GitBug-Java and SWE-Bench-Live benchmarks.

Enabled weaker models like GPT-5 and Claude Haiku 4.5 to match or surpass stronger models.

Systematic ablations confirm the importance of debugger integration and subagent architecture.

Abstract

While significant progress has been made in automating various aspects of software development through coding agents, there is still significant room for improvement in their bug fixing capabilities. Debugging and investigation of runtime behavior remains largely a manual, developer-driven process. Popular coding agents typically rely on either static analysis of the code or iterative test-fix cycles, which is akin to trial and error debugging. We posit that there is a wealth of rich runtime information that developers routinely access while debugging code, which agents are currently deprived of due to design limitations. Despite how prevalent debuggers are in modern IDEs and command-line tools, they have surprisingly not made their way into coding agents. In this work, we introduce Debug2Fix, a novel framework that incorporates interactive debugging as a core component of a software engineering agent via a subagent architecture. We incorporate debuggers for Java and Python into our agent framework and evaluate against GitBug-Java and SWE-Bench-Live and achieve >20% improvement in performance compared to the baseline for certain models. Furthermore, using our framework, we're able to make weaker models like GPT-5 and Claude Haiku 4.5 match or exceed the performances of stronger models like Claude Sonnet 4.5, showing that better tool design is often just as important as switching to a more expensive model. Finally, we conduct systematic ablations demonstrating the importance of both the subagent architecture and debugger integration.