大型语言模型（LLMs）已展现出令人印象深刻的推理能力，但其性能高度依赖于提示策略和模型规模。虽然强化学习和微调已被用于提升推理能力，但这些方法会带来大量的计算和数据开销。在本工作中，我们提出了自适应思维图（Adaptive Graph of Thoughts，AGoT），这是一种动态的、基于图的推理框架，仅在测试时增强LLM的推理能力。AGoT不依赖思维链（Chain of Thought，CoT）或思维树（Tree of Thoughts，ToT）等固定步骤方法，而是将复杂查询递归分解为结构化的子问题，形成一个由相互依赖的推理步骤组成的动态有向无环图（DAG）。通过选择性地仅扩展那些需要进一步分析的子问题，AGoT将链、树和图范式的优势统一到一个连贯的框架中，将计算分配到最需要的地方。我们在涵盖多跳检索、科学推理和数学问题求解的多个基准上验证了我们的方法，在科学推理任务（GPQA）上实现了高达46.2%的提升——与通过计算密集型强化学习方法获得的增益相当，并优于最先进的迭代方法。这些结果表明，动态分解和结构化递归为训练后修改提供了一种可扩展、成本效益高的替代方案，为LLMs中更稳健、通用的推理铺平了道路。
英文摘要：
Large Language Models (LLMs) have demonstrated impressive reasoning capabilities, yet their performance is highly dependent on the prompting strategy and model scale. While reinforcement learning and fine-tuning have been deployed to boost reasoning, these approaches incur substantial computational and data overhead. In this work, we introduce Adaptive Graph of Thoughts (AGoT), a dynamic, graph-based inference framework that enhances LLM reasoning solely at test time. Rather than relying on fixed-step methods like Chain of Thought (CoT) or Tree of Thoughts (ToT), AGoT recursively decomposes complex queries into structured subproblems, forming an dynamic directed acyclic graph (DAG) of interdependent reasoning steps. By selectively expanding only those subproblems that require further analysis, AGoT unifies the strengths of chain, tree, and graph paradigms into a cohesive framework that allocates computation where it is most needed. We validate our approach on diverse benchmarks spanning multi-hop retrieval, scientific reasoning, and mathematical problem-solving, achieving up to 46.2% improvement on scientific reasoning tasks (GPQA) - comparable to gains achieved through computationally intensive reinforcement learning approaches and outperforming state-of-the-art iterative approaches. These results suggest that dynamic decomposition and structured recursion offer a scalable, cost-effective alternative to post-training modifications, paving the way for more robust, general-purpose reasoning in LLMs.
We explore how generating a chain of thought -- a series of intermediate reasoning steps -- significantly improves the ability of large language models to perform complex reasoning. In particular, we show how such reasoning abilities emerge naturally in sufficiently large language models via a simple method called chain of thought prompting, where a few chain of thought demonstrations are provided as exemplars in prompting. Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks. Th…
Chain-of-thought prompting combined with pre-trained large language models has achieved encouraging results on complex reasoning tasks. In this paper, we propose a new decoding strategy, self-consistency, to replace the naive greedy decoding used in chain-of-thought prompting. It first samples a diverse set of reasoning paths instead of only taking the greedy one, and then selects the most consistent answer by marginalizing out the sampled reasoning paths. Self-consistency leverages the intuition that a complex reasoning problem typically admits multiple different ways of thinking leading to i…
ThoughtSource: A central hub for large language model reasoning data
Medical University of Vienna（维也纳医科大学）；Technical University of Denmark（丹麦技术大学）；Copenhagen University Hospital（哥本哈根大学医院）；The Royal Institute of Technology (KTH)（皇家理工学院（KTH））
Not All Thoughts are Generated Equal: Efficient LLM Reasoning via Multi-Turn Reinforcement Learning
AI Thrust, The Hong Kong University of Science and Technology (Guangzhou)（香港科学与技术大学（广州）人工智能方向）；Didichuxing Co. Ltd（滴滴出行有限公司）；CSE, The Hong Kong University of Science and Technology（香港科学与技术大学计算机科学与工程系）
Mitigating Misleading Chain-of-Thought Reasoning with Selective Filtering
College of Zhiyuan, Shanghai Jiao Tong University（上海交通大学致远学院）；School of Electronic Information and Electrical Engineering, Shanghai Jiao Tong University（上海交通大学电子信息与电气工程学院）；Shanghai Jiao Tong University（上海交通大学）
Distilling Mathematical Reasoning Capabilities into Small Language Models
Institute of Information Engineering, Chinese Academy of Sciences（中国科学院信息工程研究所）；School of Cyber Security, University of Chinese Academy of Sciences（中国科学院大学网络空间安全学院）；Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学高瓴人工智能学院）
AdvChain: Adversarial Chain-of-Thought Tuning for Robust Safety Alignment of Large Reasoning Models
The Chinese University of Hong Kong, Shenzhen（香港中文大学（深圳））；State University of New York at Buffalo（纽约州立大学布法罗分校）；Huawei International, Singapore（新加坡华为国际）
UniVLR: Unifying Text and Vision in Visual Latent Reasoning for Multimodal LLMs
University of Science and Technology of China（中国科学技术大学）；Tsinghua University（清华大学）；National University of Singapore（新加坡国立大学）；Zhongguancun Academy（中关村学院）
