Scaling inference compute enhances reasoning in large language models (LLMs), with long chains-of-thought (CoTs) enabling strategies like backtracking and error correction. Reinforcement learning (RL) has emerged as a crucial method for developing these capabilities, yet the conditions under which long CoTs emerge remain unclear, and RL training requires careful design choices. In this study, we systematically investigate the mechanics of long CoT reasoning, identifying the key factors that enable models to generate long CoT trajectories. Through extensive supervised fine-tuning (SFT) and RL experiments, we present four main findings: (1) While SFT is not strictly necessary, it simplifies training and improves efficiency; (2) Reasoning capabilities tend to emerge with increased training compute, but their development is not guaranteed, making reward shaping crucial for stabilizing CoT length growth; (3) Scaling verifiable reward signals is critical for RL. We find that leveraging noisy, web-extracted solutions with filtering mechanisms shows strong potential, particularly for out-of-distribution (OOD) tasks such as STEM reasoning; and (4) Core abilities like error correction are inherently present in base models, but incentivizing these skills effectively for complex tasks via RL demands significant compute, and measuring their emergence requires a nuanced approach. These insights provide practical guidance for optimizing training strategies to enhance long CoT reasoning in LLMs. Our code is available at: https://github.com/eddycmu/demystify-long-cot.
Demystifying Entropy-based Selection for Chain-of-Thought Compression in Large Reasoning Models
University of Trieste（的里雅斯特大学）；University of Milano-Bicocca（米兰-比科卡大学）；University of Groningen（格罗宁根大学）；Khoury College of Computer Sciences, Northeastern University（东北大学科里计算机学院）
Scaling Reasoning Hop Exposes Weaknesses: Demystifying and Improving Hop Generalization in Large Language Models
University of Science and Technology of China（中国科学技术大学）；City University of Hong Kong（香港城市大学）；Zhejiang University（浙江大学）；City University of Hong Kong, Shenzhen Research Institute（香港城市大学深圳研究院）
Demystifying Scientific Problem-Solving in LLMs by Probing Knowledge and Reasoning
Department of Computer Science, Yale University；Department of Molecular \& Cellular Biology, Harvard University；Allen Institute for AI；Department of Computer Science, Northwestern University
11Plus-Bench: Demystifying Multimodal LLM Spatial Reasoning with Cognitive-Inspired Analysis
Microsoft Research（微软研究院）；Language Technology Lab, University of Cambridge（语言技术实验室，剑桥大学）；Institute of Automation, Chinese Academy of Sciences（自动化研究所，中国科学院）；Department of Oncology, University of Cambridge（癌症部门，剑桥大学）；Leverhulme Centre for the Future of Intelligence, University of Cambridge（未来智能中心，剑桥大学）；VRAIN, Universitat Politècnica de València（VRAIN，巴塞罗那理工大学）
Demystifying Reasoning Dynamics with Mutual Information: Thinking Tokens are Information Peaks in LLM Reasoning
Gaoling School of Artificial Intelligence, Renmin University of China（中国人民大学北京校区人工智能学院）；Shanghai Artificial Intelligence Laboratory（上海人工智能实验室）；University College London, University of London（伦敦大学大学学院）；Dalian University of Technology（大连理工大学）
Demystifying OPD: Length Inflation and Stabilization Strategies for Large Language Models
Department of Computer Science, Rice University, Houston, USA（莱斯大学计算机科学系，美国休斯顿）；Department of Computer Science, University of North Carolina at Charlotte, Charlotte, USA（北卡罗来纳大学夏洛特分校计算机科学系，美国夏洛特）；Department of Computer Science, Johns Hopkins University, Baltimore, USA（约翰霍普金斯大学计算机科学系，美国巴尔的摩）；Department of Computer and Data Sciences, Case Western Reserve University（凯斯西储大学计算机与数据科学系）
Think Consistently, Reason Efficiently: Energy-Based Calibration for Implicit Chain-of-Thought
University of Oxford（牛津大学）；Tsinghua University（清华大学）；Tencent（腾讯）；Southern University of Science and Technology（南方科技大学）；National University of Singapore（新加坡国立大学）
Thought Graph: Generating Thought Process for Biological Reasoning
University of Texas at Austin（德克萨斯大学奥斯汀分校）；Arizona State University（亚利桑那州立大学）；University of Pennsylvania（宾夕法尼亚大学）；University of California San Diego（加利福尼亚大学圣迭戈分校）；The University of North Carolina at Chapel Hill（北卡罗来纳大学教堂山分校）；California Institute of Technology（加利福尼亚理工学院）
Towards Better Understanding of Program-of-Thought Reasoning in Cross-Lingual and Multilingual Environments
School of Information Science and Technology, VISTEC（信息科学与技术学院，VISTEC）；KAIST（韩国科学技术院）；Cohere；SCB 10X；AI Singapore（AI新加坡）；Department of Computer Engineering, Chulalongkorn University（朱拉隆梭大学计算机工程系）
Blind Spot Navigation in Large Language Model Reasoning with Thought Space Explorer
Clemson University（克莱姆森大学）；Université de Montréal（蒙特利尔大学）；Center for Advanced AI, Accenture（Accenture高级人工智能中心）；The University of Alabama（阿拉巴马大学）；University of Kansas（堪萨斯大学）；Arizona State University（亚利桑那州立大学）
