通過 MCTS 朝向 LLMs 的自我改進：利用逐步知識與課程偏好學習

Towards Self-Improvement of LLMs via MCTS: Leveraging Stepwise Knowledge with Curriculum Preference Learning

October 9, 2024

作者: Xiyao Wang, Linfeng Song, Ye Tian, Dian Yu, Baolin Peng, Haitao Mi, Furong Huang, Dong Yu

cs.AI

摘要

蒙特卡羅樹搜索（MCTS）最近已成為增強LLM推理能力的強大技術。諸如SFT或DPO等技術使LLM能夠從MCTS中提煉高質量行為，從而提高其推理性能。然而，現有的提煉方法未充分利用MCTS生成的豐富軌跡信息，限制了LLM推理能力提升的潛力。本文提出了AlphaLLM-CPL，一種新型的成對訓練框架，通過MCTS行為提煉使LLM能夠自我改進。AlphaLLM-CPL通過兩個關鍵創新有效地利用MCTS軌跡：（1）AlphaLLM-CPL從搜索樹中共享相同父節點的子節點構建逐步軌跡對，為更有效的MCTS行為提煉提供步級信息。 （2）AlphaLLM-CPL引入課程偏好學習，動態調整每個離線訓練時期中軌跡對的訓練順序，以優先考慮關鍵學習步驟並減輕過度擬合。在數學推理任務上的實驗結果表明，AlphaLLM-CPL明顯優於先前的MCTS行為提煉方法，大幅提升了LLM的推理能力。

English

Monte Carlo Tree Search (MCTS) has recently emerged as a powerful technique for enhancing the reasoning capabilities of LLMs. Techniques such as SFT or DPO have enabled LLMs to distill high-quality behaviors from MCTS, improving their reasoning performance. However, existing distillation methods underutilize the rich trajectory information generated by MCTS, limiting the potential for improvements in LLM reasoning. In this paper, we propose AlphaLLM-CPL, a novel pairwise training framework that enables LLMs to self-improve through MCTS behavior distillation. AlphaLLM-CPL efficiently leverages MCTS trajectories via two key innovations: (1) AlphaLLM-CPL constructs stepwise trajectory pairs from child nodes sharing the same parent in the search tree, providing step-level information for more effective MCTS behavior distillation. (2) AlphaLLM-CPL introduces curriculum preference learning, dynamically adjusting the training sequence of trajectory pairs in each offline training epoch to prioritize critical learning steps and mitigate overfitting. Experimental results on mathematical reasoning tasks demonstrate that AlphaLLM-CPL significantly outperforms previous MCTS behavior distillation methods, substantially boosting the reasoning capabilities of LLMs.