arXiv cs.CL· Huifei Wang, Xinying Huang, Yiheng Sun, Yifan Yuan·· 2 天前AI 评分34
ReMCTS：面向代码生成的反思增强蒙特卡洛树搜索框架
ReMCTS: Reflection-Enhanced Monte Carlo Tree Search for Code Generation
AI 导读
ReMCTS 是一个执行驱动、带记忆增强的 LLM 引导 MCTS 式搜索框架，将程序候选组织为树状态并跨分支检索失败经验。在 HumanEval 和 MBPP-Sanitized 的留出评测中，可见测试版 ReMCTS 在 10 个模型-数据集组合中的 8 个优于直接生成，而仅用代理信号搜索稳定性较差。
正文 · 原文
Abstract:Open-weight large language models (LLMs) can generate function-level programs from natural-language prompts, but plausible candidates still fail on hidden semantics and repeat mistakes across repair attempts. We present ReMCTS, an execution-grounded, memory-augmented, LLM-guided MCTS-style search framework. It organizes program candidates as tree states, retains branch-local debugging context, retrieves failure experience across branches, and distinguishes failed checks from unavailable evidence. On HumanEval and MBPP-Sanitized, visible-test ReMCTS improves over direct generation in 8 of 10 model-dataset pairs under held-out evaluation, whereas proxy-only search is less stable. Controlled tree-search, sampling, repair, and memory ablations characterize the source and limits of these gains. A 30-task HumanEval-X C++ pilot further demonstrates compatibility with compiler-backed execution, but does not constitute a broad multilingual evaluation.
| Comments: | 21 pages, 2 figures. To appear in the Proceedings of EMNLP 2026 |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2609.34717 [cs.CL] |
| (or arXiv:2609.34717v2 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2609.34717 arXiv-issued DOI via DataCite |
Submission history
From: Huifei Wang [view email]
[v1]
Mon, 28 Sep 2026 09:24:20 UTC (640 KB)
[v2]
Tue, 27 Oct 2026 03:00:48 UTC (640 KB)
来源：arXiv cs.CL · arxiv.org