在LLMs的测试时间扩展中对树搜索策略与固定令牌预算对齐
Aligning Tree-Search Policies with Fixed Token Budgets in Test-Time Scaling of LLMs
AI总结：
本文提出了一种名为Budget-Guided MCTS (BG-MCTS)的树搜索解码算法，通过将搜索策略与剩余令牌预算对齐，以提高在不同令牌预算下的推理性能。
摘要：
树搜索解码是大型语言模型（LLMs）测试时间扩展的有效方法，但现实部署中通常会施加一个固定的每查询令牌预算，且该预算在不同设置中有所不同。现有的树搜索策略大多缺乏预算意识，仅将预算视为终止条件，从而可能导致后期过度分支或提前终止。我们提出Budget-Guided MCTS (BG-MCTS)，一种树搜索解码算法，其搜索策略与剩余令牌预算对齐：它从广泛的探索开始，然后在剩余预算减少时优先进行细化和答案完成，同时减少浅层节点的后期分支。BG-MCTS在数学推理基准和额外的物理推理基准上，使用开放权重LLMs在各种推理预算下均优于预算无关的树搜索基线。
英文摘要：
Tree-search decoding is an effective form of test-time scaling for large language models (LLMs), but real-world deployment often imposes a fixed per-query token budget that varies across settings. Existing tree-search policies are largely budget-agnostic, treating the budget merely as a termination condition, thereby risking late-stage over-branching or premature termination. We propose Budget-Guided MCTS (BG-MCTS), a tree-search decoding algorithm that aligns its search policy with the remaining token budget: it starts with broad exploration, then prioritizes refinement and answer completion as the remaining budget decreases while reducing late-stage branching from shallow nodes. BG-MCTS consistently outperforms budget-agnostic tree-search baselines across inference budgets on mathematical reasoning benchmarks and an additional physics reasoning benchmark with open-weight LLMs.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
