Verbal Process Supervision Elicits Better Coding Agents
Hao-Yuan Chen, Cheng-Pong Huang, Jui-Ming Yao
2025-03-25
Summary
This paper is about making AI coding assistants better by having them explain their thought process while they code.
What's the problem?
Even the best AI coding assistants sometimes struggle with complex coding tasks.
What's the solution?
The researchers developed a new system called CURA that makes the AI verbally explain its reasoning as it writes code. This helps the AI stay on track and produce better code.
Why it matters?
This work matters because it shows that having AI explain its reasoning can significantly improve its coding abilities, which can make software development faster and easier.
Abstract
The emergence of large language models and their applications as AI agents have significantly advanced state-of-the-art code generation benchmarks, transforming modern software engineering tasks. However, even with test-time computed reasoning models, these systems still struggle with complex software engineering challenges. This work introduces CURA, a code understanding and reasoning agent system enhanced with verbal process supervision (VPS), achieving a 3.65\% improvement over baseline models on challenging benchmarks like BigCodeBench. Furthermore, CURA, when paired with the o3-mini model and VPS techniques, attains state-of-the-art performance. This work represents a step forward in integrating reasoning-driven architectures with LLM-based code generation, enabling agentic reasoning for language models to solve complex software engineering tasks.