让大模型自适应选择推理策略，无需手动设计提示词。
Mixture of Reasonings: Teach Large Language Models to Reason with Adaptive Strategies
- 训练模型掌握多种推理路径，自动选择最优解法。
- 在基准测试中准确率提升13.5%，优于基线方法。
- 适合需要灵活推理的复杂任务场景，如数学与逻辑题。
大型语言模型（LLMs）通过链式思维（CoT）和思维树（ToT）等高级提示技术在复杂任务中表现优异，但其对人工设计的任务特定提示的依赖限制了适应性和效率。我们提出混合推理（Mixture of Reasoning, MoR）训练框架，将多种推理策略嵌入LLMs中，实现无需外部提示工程的自主、任务自适应推理。MoR包含两个阶段：思维生成阶段，利用GPT-4o等模型生成推理链模板；SFT数据集构建阶段，将模板与基准数据集配对用于监督微调。实验表明，MoR显著提升性能，其中MoR150在使用CoT提示时达到0.730（相比基线提升2.2%），在其他设置下达0.734（提升13.5%）。该方法消除了对任务特定提示的需求，为多样化任务提供了可泛化的强健推理解决方案。
原文摘要 · Abstract (English)
Large language models (LLMs) excel in complex tasks through advanced prompting techniques like Chain-of-Thought (CoT) and Tree-of-Thought (ToT), but their reliance on manually crafted, task-specific prompts limits adaptability and efficiency. We introduce Mixture of Reasoning (MoR), a training framework that embeds diverse reasoning strategies into LLMs for autonomous, task-adaptive reasoning without external prompt engineering. MoR has two phases: Thought Generation, creating reasoning chain templates with models like GPT-4o, and SFT Dataset Construction, pairing templates with benchmark datasets for supervised fine-tuning. Our experiments show that MoR significantly enhances performance, with MoR150 achieving 0.730 (2.2% improvement) using CoT prompting and 0.734 (13.5% improvement) compared to baselines. MoR eliminates the need for task-specific prompts, offering a generalizable solution for robust reasoning across diverse tasks.
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。