@inproceedings{li-etal-2025-codetree,
    title = "{C}ode{T}ree: Agent-guided Tree Search for Code Generation with Large Language Models",
    author = "Li, Jierui  and
      Le, Hung  and
      Zhou, Yingbo  and
      Xiong, Caiming  and
      Savarese, Silvio  and
      Sahoo, Doyen",
    editor = "Chiruzzo, Luis  and
      Ritter, Alan  and
      Wang, Lu",
    booktitle = "Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 1: Long Papers)",
    month = apr,
    year = "2025",
    address = "Albuquerque, New Mexico",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/fix-sig-urls/2025.naacl-long.189/",
    pages = "3711--3726",
    ISBN = "979-8-89176-189-6",
    abstract = "Pretrained 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, achieving a 31.9{\%} solving rate."
}