Reasoning-capable large language models solve hard problems by emitting long chains of thought, paying heavily in latency, GPU time, and energy. Casual inspection of their traces reveals extensive reformulation, verification, and circular self-reflection, yet how much of this deliberation is actually necessary has never been measured at scale or explained from first principles. This paper closes both gaps.
We formalise reasoning redundancy directly in terms of the reasoning model itself: the redundancy of a correct trace is the largest fraction of its trailing segmented steps that can be truncated while $π$, forced to terminate thinking and emit a final answer, still produces the correct answer. A large-scale quantification across four frontier reasoning models and two mathematical benchmarks shows that step-level redundancy is consistently high -- between 61% and 93% across the 8 (model, benchmark) conditions we study, with the median critical prefix equal to a single segmented step in six of the eight conditions -- that the finding is robust to the choice of judge family, and that although $ρ$ decreases with problem difficulty on MATH-500, all four models remain substantially redundant ($ρ\in [46\%, 85\%]$) even on the hardest Level-5 problems.
We then prove that this redundancy is a structural consequence of length-agnostic outcome rewards, not a model-specific artefact: under any such reward, no finite expected stopping time is optimal. The result holds regardless of RL algorithm, base model, data distribution, or whether the policy is obtained via RL or distillation; over-thinking is therefore not a bug to be patched in individual models but a structural property of how current reasoning models are trained. Code: https://github.com/zhiyuanZhai20/how-much-thinking-is-enough
Harbin Institute of Technology (Shenzhen)(哈尔滨工业大学(深圳));Huawei Noah’s Ark Lab(华为诺亚方舟实验室);Tsinghua University(清华大学);The Chinese University of Hong Kong(香港中文大学);Zhongguancun Academy(中关村学院);Shenzhen Loop Area Institute(深圳河套学院)
Beyond Frequency: The Role of Redundancy in Large Language Model Memorization
Semyung University(_semyung大学);University of Copenhagen(哥本哈根大学);University of Washington(华盛顿大学);SAIBD, Hefei University(合肥大学);School of Information Science and Engineering, Lanzhou University(兰州大学信息科学与工程学院)
Ontology-Guided Reverse Thinking Makes Large Language Models Stronger on Knowledge Graph Question Answering
Harbin Institute of Technology(哈尔滨工业大学);Beijing University of Posts and Telecommunications(北京邮电大学);Joint Laboratory of HIT and iFLYTEK(哈工大与科大讯飞联合实验室);University of Science and Technology of China(中国科学技术大学);Pengcheng Laboratory(鹏城实验室)
The Evolution of Thought: Tracking LLM Overthinking via Reasoning Dynamics Analysis
State Key Laboratory of AI Safety,Institute of Computing Technology, Chinese Academy of Sciences(人工智能安全国家重点实验室,计算技术研究所,中国科学院);University of Chinese Academy of Sciences(中国科学院大学)
MedPruner: Training-Free Hierarchical Token Pruning for Efficient 3D Medical Image Understanding in Vision-Language Models
Chinese University of Hong Kong(香港中文大学);Westlake University(西湖大学);Southern Medical University(南方医科大学);Jiangnan University(江南大学);Dalian University of Technology(大连理工大学);Shanghai Jiao Tong University(上海交通大学);University of Copenhagen(哥本哈根大学)
Cognitive Mismatch in Multimodal Large Language Models for Discrete Symbol Understanding
Tsinghua University(清华大学);Sun Yat-sen University(中山大学);Independent Researcher(独立研究员);The Hong Kong Polytechnic University(香港理工大学);Guangdong Laboratory of AI and Digital Economy (SZ)(广东省人工智能与数字经济实验室(深圳));University of Illinois Chicago(伊利诺伊大学芝加哥分校)
LookWise: Knowing When and Where to Look for Fine-Grained Visual Reasoning in Multimodal Large Language Models
Institute of Intelligent Machines, Hefei Institutes of Physical Science, Chinese Academy of Sciences(智能机器研究所,合肥物理科学研究院,中国科学院);University of Science and Technology of China(中国科学技术大学);Zhejiang University(浙江大学);East China Normal University(华东师范大学);The Hong Kong University of Science and Technology(香港科技大学)
You Only Need 2/3 of the Chosen Experts: An Empirical Study of Dynamic Expert Pruning in Fine-Grained MoE LLMs
Institute of Automation, Chinese Academy of Sciences(中国科学院自动化研究所);School of Artificial Intelligence, University of Chinese Academy of Sciences(中国科学院大学人工智能学院);Zhongguancun Academy(中关村学院);University of Science and Technology of China(中国科学技术大学)
Bridging Modal Isolation in Interleaved Thinking: Supervising Modality Transitions via Stepwise Reinforcement
Shanghai Artificial Intelligence Laboratory(上海人工智能实验室);Shanghai Jiaotong University(上海交通大学);Zhejiang University(浙江大学);University of Chinese Academy of Sciences(中国科学院大学)