CoT Compression

CoT: Chain-of-Thought

Chain-of-thought compression studies how to make language models reason with fewer generated tokens or less computation without losing the information needed to solve a task. Approaches include pruning low-utility tokens or segments, reinforcement learning with adaptive length rewards, distilling concise traces, and replacing explicit steps with latent representations or reusable memory. Methods are commonly tested on mathematical reasoning benchmarks such as GSM8K, MATH, and AIME, as well as science, general-knowledge, and agent tasks. The central tension is that shorter traces can preserve accuracy in many settings, but aggressive or poorly targeted compression may harm difficult reasoning, reliability, or trustworthiness.

Latest papers

57 papers7 papers in the last four weeks, up 75% on the four weeks before. 0.0% of all new papers.

Improving the Chain-of-Thought Compressibility of Pruned Reasoning Models via Long--Short Contrastive Pruning
Oct 25, 2026Tao Jing, Chen Kang, Feifei Zhai +2Efficient Language Model Reasoning LLM Pruning
Learning to Reason Efficiently with Interventional Step Credit
Oct 19, 2026Guanda Li, Xuefeng Bai, Yulong Chen +3Efficient Language Model Reasoning Step-Level Credit Assignment in RL
Harbin Institute of Technology, Shenzhen · University of Aberdeen · University of Cambridge
Think, Compress, Answer: Reinforcement Learning for Multimodal Grounding via Reasoning Compression
Oct 19, 2026Jinbae Seo, Hyeongjun Kwon, Jiyoung Lee +4Multimodal Grounding Multimodal Large Language Models
Yonsei University · School of AI and Software, Ewha Womans University · LIG Defense & Aerospace +1
Does the Linguistic Register of Thinking Matter? Wenyan as a Compact Reasoning Register for Chinese LLMs
Oct 14, 2026Minghan Li, Guodong ZhouEfficient Language Model Reasoning Language Modeling
Soochow University, Suzhou, China
DoubtLess: Training-Free Contrastive Decoding for Efficient Reasoning
Oct 13, 2026Minghao Luo, Liang ChenEfficient Language Model Reasoning CoT Compression
The Chinese University of Hong Kong · EPFL
Quantifying the Stability of Multi-Step Reasoning via Error Amplification
Oct 5, 2026Dongyue Li, Ziniu Zhang, Minxuan Duan +1Efficient Language Model Reasoning CoT Compression
Northeastern University Boston, MA
SYNLAT: Syntax-Aligned Text-Latent Compression for Chain-of-Thought Reasoning
Oct 2, 2026Yifeng Zhao, Hongjun Yu, Shibo Wang +5CoT Compression Efficient Language Model Reasoning
Nanyang Technological University · Nankai University
On the Token Value Inequality in Efficient Reasoning
Sep 27, 2026Runjia Zeng, Hang Hua, Yiyang Liu +5RL for Language Model Reasoning Efficient Language Model Reasoning
Rochester Institute of Technology · Purdue University · MIT-IBM Watson AI Lab +3
When Can Agents Forget Their Reasoning? ICLR for Long-Horizon Agent Context Compression
Sep 24, 2026Mingxuan Wang, Fei Luo, Bo Wang +6Long-Context Modeling Long-Horizon LLM Agents
TierFlow Team · Gaoling School of Artificial Intelligence, Renmin University of China · Tsinghua University
A*-Thought-V2: Efficient Latent Reasoning via Geometric Dynamics of LLM
Sep 7, 2026Xiaoang Xu, Siyuan Liu, Shuo Wang +13Efficient Language Model Reasoning Continuous Latent Reasoning
Beijing University of Posts and Telecommunications · OpenBMB · The Hong Kong Polytechnic University +2
Every Token Leaves a Ripple in the Stream of Thought: Eliciting Model-Internal Token Saliency for Chain-of-Thought Compression
Aug 31, 2026Tianyi Zhao, Yinhan He, Wendy Zheng +2LLM Compression Efficient Language Model Reasoning
University of Virginia
ChainPrune: Evaluating and Reducing Redundancy in Long Chain-of-Thought Reasoning
Aug 22, 2026Weihang Pan, Zhengxu Yu, Yuxiang Zhang +5Efficient Language Model Reasoning Overthinking in Language Models
Zhejiang University · Alibaba Group
Memory Augmentation Unlocks Efficient Chain-of-Thought Reasoning
Aug 21, 2026Simeng Zhang, Yilong Chen, Wenyuan Zhang +4Efficient Language Model Reasoning CoT Compression
Institute of Information Engineering, Chinese Academy of Sciences · School of Cyber Security, University of Chinese Academy of Sciences · Tencent Inc. +1
Demystifying Entropy-based Selection for Chain-of-Thought Compression in Large Reasoning Models
Jul 30, 2026Sara Candussio, Daniel Scalena, Luca Bortolussi +3CoT Compression Large Reasoning Models
University of Trieste · University of Milano-Bicocca · CLCG, University of Groningen +2
Efficient Chain-of-Modality Reasoning via Progressive Compression for Spoken Language Models
Jul 22, 2026Pengchao Feng, Chao-Hong Tan, Qian Chen +3Efficient Language Model Reasoning Speech Language Models
Shanghai Jiao Tong University · Shanghai Innovation Institute · Shanghai, China +2
Training Continuous Chain of Thought Models: A Tale of Two Regimes
Jul 18, 2026Varun Yerram, He He, Eunsol ChoiCoT Reasoning Continuous Latent Reasoning
Center for Data Science, New York University
Valid $\ne$ Necessary: Diagnosing Latent Inefficiency in Chain-of-Thought
Jul 13, 2026Daeyeop Lee, Hwanjo YuLLM Inference Efficiency Overthinking in Language Models
1KT Corporation · 2Pohang University of Science and Technology
When Compression Helps and When It Hurts: Condition-Aware Analysis of Chain-of-Thought Distillation
Jun 19, 2026Siyang Lyu, Xinghao Chen, Zhijing Sun +3Model Compression CoT Distillation
Ningbo Institute of Digital Twin, Eastern Institute of Technology, Ningbo · Viterbi School of Engineering, University of Southern California · The Hong Kong Polytechnic University +2
Masked Self-Distillation: Internalizing the Chain-of-Thought in Language Models
Jun 18, 2026Durgesh Kalwar, Vardhan Palod, Jaya Adithya Pavuluri +1CoT Distillation Efficient Language Model Reasoning
SCAI, Arizona State University
SuperThoughts: Reasoning Tokens in Superposition
Jun 11, 2026Zheyang Xiong, Shivam Garg, Max Yu +4Multi-Token Prediction LLM Inference Acceleration
mMicrosoft Research · iIndependent · pPrinceton University +2
ReSum: Synergizing LLM Reasoning and Summarization with Reinforcement Learning
Jun 11, 2026Xucong Wang, Ziyu Ma, Yong Wang +5RL for Language Model Reasoning Reinforcement Learning with Verifiable Rewards
University of Science and Technology of China · 2AMAP, Alibaba Group
Compress-Distill: Reasoning Trace Compression for Efficient Knowledge Distillation
Jun 4, 2026Maxime Griot, Paul Steven Scotti, Tanishq Mathew AbrahamCoT Distillation Teacher-Student Learning
Université catholique de Louvain · Sophont Inc.
HybridThinker: Efficient Chain-of-Thought Reasoning via Compressed Memory and Transient Thought Steps
Jun 2, 2026Xin Liu, Runsong Zhao, Xinyu Liu +8Memory-Augmented Language Models Efficient Language Model Reasoning
School of Computer Science and Engineering, Northeastern University, Shenyang 110819, China · China Unicom Cloud-Link, China · NiuTrans Research, Shenyang, China
ThoughtFold: Folding Reasoning Chains via Introspective Preference Learning
Jun 2, 2026Ziyan Liu, Xueda Shen, Yuzhe Gu +7Efficient Language Model Reasoning Preference Optimization
Shanghai Artificial Intelligence Laboratory · University of Science and Technology of China · MMLab, The Chinese University of Hong Kong
HMPO: Hybrid Median-length Policy Optimization for Chain-of-Thought Compression
Jun 1, 2026Minghui Zheng, Hongxu Chen, Huimin Ren +8LLM Compression RL for Language Model Reasoning
Li Auto Inc.
SLAT: Segment-Level Adaptive Trimming for Efficient CoT Reasoning
May 29, 2026Jian Yao, Xiongcai Luo, Ran Cheng +1RL for Language Model Reasoning Efficient Language Model Inference
Department of Data Science and Artificial Intelligence, The Hong Kong Polytechnic University · Bytedance · The Hong Kong Polytechnic University-Daya Bay Technology and Innovation Research Institute, Huizhou,Guangdong Province, China.
Zipping the Thought: When and How Compressed Reasoning Data Works in LLM Post-Training
May 27, 2026Kohsei Matsutani, Gouki Minegishi, Takeshi Kojima +2Supervised Fine-Tuning LLM Post-Training
The University of Tokyo
Selective Latent Thinking: Adaptive Compression of LLM Reasoning Chains
May 25, 2026Hui Xie, Jie Liu, Ziyue Qiao +1Efficient Language Model Reasoning LLM Inference Acceleration
Eindhoven University of Technology, Netherlands · School of Computing and Information Technology, Great Bay University, China
Uncovering the Representation Geometry of Minimal Cores in Overcomplete Reasoning Traces
May 14, 2026Sanjoy Chowdhury, Dinesh ManochaNeural Representation Geometry LLM Interpretability
University of Maryland, College Park, USA
RuPLaR : Efficient Latent Compression of LLM Reasoning Chains with Rule-Based Priors From Multi-Step to One-Step
May 10, 2026Xiaocheng Luo, Kang Wang, Zaifu Zhan +2LLM Compression CoT Compression
School of Computer Science and Technology, Soochow University, Suzhou, 215006, Jiangsu, PR China · Department of Electrical and Computer Engineering, University of Minnesota, Minneapolis, 55455, MN, USA

Oct 25, 2026J/K move · Enter open · S save

Improving the Chain-of-Thought Compressibility of Pruned Reasoning Models via Long--Short Contrastive Pruning

Tao Jing, Chen Kang, Feifei Zhai, Pengyuan Liu+1

Efficient Language Model Reasoning LLM Pruning CoT Compression Neural Network Pruning

Open paper AskPDF arXiv

Large reasoning models are costly to serve because they are large and because they generate long chains of thought (CoT). Weight pruning and CoT compression address these two costs separately, but they interfere in that pruning adds redundant reasoning on problems the model still solves and incomplete reasoning on those it no longer solves, and on harder problems makes the model less compressible by subsequent length-preference optimization. We trace this to a blind spot on each side: pruning criteria do not distinguish necessary from redundant reasoning, and preference-based CoT compression receives no training signal on problems a pruned model always or never solves. We propose Contrastive Set-Difference (CSD), which identifies neurons enriched in redundant reasoning by contrasting their activations on long and short correct solutions of the same problem and prunes them preferentially, and we extend preference pairs to the problems a pruned model always or never solves. Pruning 20% of DeepSeek-R1-Distill-Qwen-7B, the combination uses 8--22% fewer thinking tokens on four math benchmarks than pruning with reasoning-aware calibration (RAC) followed by the same compression, with accuracy within 0.5 points.