提出新方法提升大模型在长文本中多步推理的准确性
Reasoning on Multiple Needles In A Haystack
- 将思考过程拆解为检索与推理阶段，引入多轮反思机制
- 发现上下文越长模型推理步骤越短，导致准确率下降
- 可提升GPT-4o在AIME2024数学题上的表现，适合长文本推理研究
Needle In A Haystack（NIAH）任务被广泛用于评估大语言模型（LLMs）处理长上下文问答的能力。然而，其依赖简单检索的方法限制了评估效果。为此，近期研究提出了多重针入草堆推理（MNIAH-R）任务，将多跳推理所需的多份支持文档（多重针）嵌入到干扰性上下文（草堆）中。尽管如此，现有方法仍无法解决模型直接从内部知识中给出答案的问题，且未解释或缓解准确率随上下文长度增加而下降的现象。本文通过过滤直接回答类问题，揭示性能下降主要源于输入长度增加导致推理过程变短。基于此，我们将思考过程分解为检索与推理两个阶段，并引入多轮反思机制。此外，我们使用生成的迭代推理过程训练模型，有效缓解了性能退化。最后，我们展示了该检索-反思能力在数学推理场景中的应用，显著提升了GPT-4o在AIME2024上的表现。
原文摘要 · Abstract (English)
The Needle In A Haystack (NIAH) task has been widely used to evaluate the long-context question-answering capabilities of Large Language Models (LLMs). However, its reliance on simple retrieval limits its effectiveness. To address this limitation, recent studies have introduced the Multiple Needles In A Haystack Reasoning (MNIAH-R) task, which incorporates supporting documents (Multiple needles) of multi-hop reasoning tasks into a distracting context (Haystack}). Despite this advancement, existing approaches still fail to address the issue of models providing direct answers from internal knowledge, and they do not explain or mitigate the decline in accuracy as context length increases. In this paper, we tackle the memory-based answering problem by filtering out direct-answer questions, and we reveal that performance degradation is primarily driven by the reduction in the length of the thinking process as the input length increases. Building on this insight, we decompose the thinking process into retrieval and reasoning stages and introduce a reflection mechanism for multi-round extension. We also train a model using the generated iterative thinking process, which helps mitigate the performance degradation. Furthermore, we demonstrate the application of this retrieval-reflection capability in mathematical reasoning scenarios, improving GPT-4o's performance on AIME2024.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。