Towards a Neural Debugger for Python

Researchers have developed neural debuggers - AI models that can emulate traditional Python debuggers by stepping through code execution, setting breakpoints, and predicting both forward and backward program states. This breakthrough enables more interactive control over neural code interpretation compared to existing approaches that only execute programs linearly.

• →Neural debuggers can perform traditional debugging operations like stepping into functions and setting breakpoints on specific code lines.
• →The models support both forward execution prediction and inverse execution inference of program states.
• →Fine-tuned large language models and smaller pre-trained models both demonstrated reliable debugging capabilities.
• →Strong performance was achieved on CruxEval benchmarks for both output and input prediction tasks.
• →This technology could enable future AI coding systems to use neural debuggers as world models for simulated debugging environments.