CodeTree: Agent-guided Tree Search for Code Generation with Large Language Models

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Li, Jierui, Le, Hung, Zhou, Yingbo, Xiong, Caiming, Savarese, Silvio, Sahoo, Doyen
Format:	Preprint
Published:	2024
Subjects:	Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

_version_	1866917835256627200
author	Li, Jierui Le, Hung Zhou, Yingbo Xiong, Caiming Savarese, Silvio Sahoo, Doyen
author_facet	Li, Jierui Le, Hung Zhou, Yingbo Xiong, Caiming Savarese, Silvio Sahoo, Doyen
contents	Pre-trained on massive amounts of code and text data, large language models (LLMs) have demonstrated remarkable achievements in performing code generation tasks. With additional execution-based feedback, these models can act as agents with capabilities to self-refine and improve generated code autonomously. However, on challenging coding tasks with extremely large search space, current agentic approaches still struggle with multi-stage planning, generating, and debugging. To address this problem, we propose CodeTree, a framework for LLM agents to efficiently explore the search space in different stages of the code generation process. Specifically, we adopted a unified tree structure to explicitly explore different coding strategies, generate corresponding coding solutions, and subsequently refine the solutions. In each stage, critical decision-making (ranking, termination, expanding) of the exploration process is guided by both the environmental execution-based feedback and LLM-agent-generated feedback. We comprehensively evaluated CodeTree on 7 code generation benchmarks and demonstrated the significant performance gains of CodeTree against strong baselines. Using GPT-4o as the base model, we consistently achieved top results of 95.1 on HumanEval, 98.7 on MBPP, and 43.0 on CodeContests. On the challenging SWEBench benchmark, our approach led to significant performance gains.
format	Preprint
id	arxiv_https___arxiv_org_abs_2411_04329
institution	arXiv
publishDate	2024
record_format	arxiv
spellingShingle	CodeTree: Agent-guided Tree Search for Code Generation with Large Language Models Li, Jierui Le, Hung Zhou, Yingbo Xiong, Caiming Savarese, Silvio Sahoo, Doyen Computation and Language Pre-trained on massive amounts of code and text data, large language models (LLMs) have demonstrated remarkable achievements in performing code generation tasks. With additional execution-based feedback, these models can act as agents with capabilities to self-refine and improve generated code autonomously. However, on challenging coding tasks with extremely large search space, current agentic approaches still struggle with multi-stage planning, generating, and debugging. To address this problem, we propose CodeTree, a framework for LLM agents to efficiently explore the search space in different stages of the code generation process. Specifically, we adopted a unified tree structure to explicitly explore different coding strategies, generate corresponding coding solutions, and subsequently refine the solutions. In each stage, critical decision-making (ranking, termination, expanding) of the exploration process is guided by both the environmental execution-based feedback and LLM-agent-generated feedback. We comprehensively evaluated CodeTree on 7 code generation benchmarks and demonstrated the significant performance gains of CodeTree against strong baselines. Using GPT-4o as the base model, we consistently achieved top results of 95.1 on HumanEval, 98.7 on MBPP, and 43.0 on CodeContests. On the challenging SWEBench benchmark, our approach led to significant performance gains.
title	CodeTree: Agent-guided Tree Search for Code Generation with Large Language Models
topic	Computation and Language
url	https://arxiv.org/abs/2411.04329