[2602.18571] Debug2Fix: Supercharging Coding Agents with Interactive Debugging Capabilities

[2602.18571] Debug2Fix: Supercharging Coding Agents with Interactive Debugging Capabilities

arXiv - AI 4 min read Article

Summary

The paper introduces Debug2Fix, a framework enhancing coding agents with interactive debugging capabilities, improving bug-fixing performance significantly.

Why It Matters

As software development increasingly relies on automation, enhancing coding agents with better debugging tools is crucial. Debug2Fix addresses existing limitations in coding agents, potentially transforming how developers interact with debugging processes and improving overall software quality.

Key Takeaways

  • Debug2Fix integrates interactive debugging into coding agents.
  • The framework shows over 20% performance improvement in bug fixing.
  • Weaker models can match stronger ones with better tool design.
  • Incorporating debuggers is essential for enhancing coding agent capabilities.
  • Systematic ablations highlight the importance of subagent architecture.

Computer Science > Software Engineering arXiv:2602.18571 (cs) [Submitted on 20 Feb 2026] Title:Debug2Fix: Supercharging Coding Agents with Interactive Debugging Capabilities Authors:Spandan Garg, Yufan Huang View a PDF of the paper titled Debug2Fix: Supercharging Coding Agents with Interactive Debugging Capabilities, by Spandan Garg and 1 other authors View PDF HTML (experimental) 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 perfor...

Related Articles

Ai Agents

AI agents have been blindly guessing your UI this whole time. Here's the file that fixes it.

Every time you ask an AI coding agent to build UI, it invents everything from scratch. Colors. Fonts. Spacing. Button styles. All of it -...

Reddit - Artificial Intelligence · 1 min ·
Llms

OpenClaw security checklist: practical safeguards for AI agents

Here is one of the better quality guides on the ensuring safety when deploying OpenClaw: https://chatgptguide.ai/openclaw-security-checkl...

Reddit - Artificial Intelligence · 1 min ·
Machine Learning

Auto agent - Self improving domain expertise agent

someone opensource an ai agent that autonomously upgraded itself to #1 across multiple domains in < 24 hours…. then open sourced the e...

Reddit - Artificial Intelligence · 1 min ·
Walmart CEO reportedly brags that company's in-app AI agent is making people spend 35% more money
Nlp

Walmart CEO reportedly brags that company's in-app AI agent is making people spend 35% more money

AI Tools & Products · 4 min ·
More in Ai Agents: This Week Guide Trending

No comments

No comments yet. Be the first to comment!

Stay updated with AI News

Get the latest news, tools, and insights delivered to your inbox.

Daily or weekly digest • Unsubscribe anytime