Authors: Kyla Levin, Nicolas van Kempen, Emery D. Berger, Stephen N. Freund

Published on: March 25, 2024

Impact Score: 8.0

Arxiv code: Arxiv:2403.16354

Summary

• What is new: Introduces ChatDBG, the first AI-powered debugging assistant that uses large language models to engage in collaborative dialogue with programmers for debugging.
• Why this is important: Existing debuggers lack the capability to engage in complex, open-ended dialogue with programmers, making root cause analysis and bug-fixing cumbersome.
• What the research proposes: ChatDBG uses large language models to allow programmers to interact with the debugger in a conversational manner, enabling it to autonomously navigate through code, analyze root causes, and suggest fixes.
• Results: ChatDBG can successfully analyze and fix a wide range of real-world errors with high accuracy. For Python programs, it led to an actionable bug fix 67% of the time with one query, which increased to 85% with one additional follow-up query.

Technical Details

Technological frameworks used: Integrates with LLDB, GDB, WinDBG, and Pdb

Models used: Large language models

Data used: Diverse set of C/C++ and Python code, including standalone scripts and Jupyter notebooks

Potential Impact

This innovation has the potential to disrupt the market for debugging tools and services, benefiting software development companies and programmers by streamlining the debugging process.

Want to implement this idea in a business?

We have generated a startup concept here: CodeCompanion.