Verbal Process Supervision Elicits Better Coding Agents

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Chen, Hao-Yuan, Huang, Cheng-Pong, Yao, Jui-Ming
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912290283978752
author Chen, Hao-Yuan
Huang, Cheng-Pong
Yao, Jui-Ming
author_facet Chen, Hao-Yuan
Huang, Cheng-Pong
Yao, Jui-Ming
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2503_18494
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Verbal Process Supervision Elicits Better Coding Agents
Chen, Hao-Yuan
Huang, Cheng-Pong
Yao, Jui-Ming
Artificial Intelligence
Computation and Language
Machine Learning
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.
title Verbal Process Supervision Elicits Better Coding Agents
topic Artificial Intelligence
Computation and Language
Machine Learning
url https://arxiv.org/abs/2503.18494