arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2604.01413cs.CLcs.AI
收藏

多轮LLM推理中的自适应停止

Adaptive Stopping for Multi-Turn LLM Reasoning

University of Illinois Chicago(伊利诺伊大学芝加哥分校);Augustana College(奥古斯塔纳学院);University of Utah(犹他大学)

Xiaofan Zhou, Huy Nguyen, Bo Yu, Chenxi Liu, Lu Cheng

AI总结:

本文提出MiCP框架,通过分配不同误差预算实现多轮推理中的自适应停止,同时保证覆盖率,减少轮次、推理成本和预测集规模。

摘要:

大型语言模型(LLMs)越来越多地依赖多轮推理和交互,如自适应检索增强生成(RAG)和ReAct风格代理,以回答难题。这些方法通过迭代检索、推理或行动提高准确性,但引入了关键挑战:何时应停止?现有方法依赖启发式停止规则或固定轮次预算,无法保证最终预测包含正确答案。本文提出多轮语言模型与置信预测(MiCP),首个支持多轮推理的置信预测框架。MiCP在不同轮次分配误差预算,使模型能提前停止同时保持整体覆盖率。在自适应RAG和ReAct上,MiCP在单跳和多跳问答基准上实现目标覆盖率,减少轮次、推理成本和预测集规模,并引入新指标联合评估覆盖率有效性与回答效率。

英文摘要:

Large Language Models (LLMs) increasingly rely on multi-turn reasoning and interaction, such as adaptive retrieval-augmented generation (RAG) and ReAct-style agents, to answer difficult questions. These methods improve accuracy by iteratively retrieving information, reasoning, or acting, but introduce a key challenge: \textbf{When should the model stop?} Existing approaches rely on heuristic stopping rules or fixed turn budgets and provide no formal guarantees that the final prediction still contains the correct answer. This limitation is particularly problematic in high-stakes domains such as finance and healthcare, where unnecessary turns increase cost and latency, while stopping too early risks incorrect decisions. Conformal prediction (CP) provides formal coverage guarantees, but existing LLM-CP methods only apply to a single model output and cannot handle multi-turn pipelines with adaptive stopping. To address this gap, we propose Multi-Turn Language Models with Conformal Prediction (MiCP), the first CP framework for multi-turn reasoning. MiCP allocates different error budgets across turns, enabling the model to stop early while maintaining an overall coverage guarantee. We demonstrate MiCP on adaptive RAG and ReAct, where it achieves the target coverage on both single-hop and multi-hop question answering benchmarks while reducing the number of turns, inference cost, and prediction set size. We further introduce a new metric that jointly evaluates coverage validity and answering efficiency.

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。

↑