Image de couverture de livre

SmartThinker: Learning to Compress and Preserve Reasoning by Step-Level Length Control

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	He, Xingyang, Ling, Xiao, Liu, Jie
Format:	Preprint
Publié:	2025
Sujets:	Artificial Intelligence Computation and Language
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

SmartThinker: Progressive Chain-of-Thought Length Calibration for Efficient Large Language Model Reasoning
par: Hu, Chenzhi, et autres
Publié: (2026)

LightThinker: Thinking Step-by-Step Compression
par: Zhang, Jintian, et autres
Publié: (2025)

On the Step Length Confounding in LLM Reasoning Data Selection
par: Wang, Bing, et autres
Publié: (2026)

LightThinker++: From Reasoning Compression to Memory Management
par: Zhu, Yuqi, et autres
Publié: (2026)

Optimizing Length Compression in Large Reasoning Models
par: Cheng, Zhengxiang, et autres
Publié: (2025)

The Impact of Reasoning Step Length on Large Language Models
par: Jin, Mingyu, et autres
Publié: (2024)

ConfSpec: Efficient Step-Level Speculative Reasoning via Confidence-Gated Verification
par: Liu, Siran, et autres
Publié: (2026)

C2F-Thinker: Coarse-to-Fine Reasoning with Hint-Guided Reinforcement Learning for Multimodal Sentiment Analysis
par: Luo, Miaosen, et autres
Publié: (2026)

KAG-Thinker: Interactive Thinking and Deep Reasoning in LLMs via Knowledge-Augmented Generation
par: Zhang, Dalong, et autres
Publié: (2025)

ProxyThinker: Test-Time Guidance through Small Visual Reasoners
par: Xiao, Zilin, et autres
Publié: (2025)

Leaky Thoughts: Large Reasoning Models Are Not Private Thinkers
par: Green, Tommaso, et autres
Publié: (2025)

The Markovian Thinker: Architecture-Agnostic Linear Scaling of Reasoning
par: Aghajohari, Milad, et autres
Publié: (2025)

Prompt-Based Length Controlled Generation with Multiple Control Types
par: Jie, Renlong, et autres
Publié: (2024)

ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoning
par: Qiao, Ziqing, et autres
Publié: (2025)

Semantic Integrity Matters: Benchmarking and Preserving High-Density Reasoning in KV Cache Compression
par: Liu, Xiang, et autres
Publié: (2025)

Learn to Reason Efficiently with Adaptive Length-based Reward Shaping
par: Liu, Wei, et autres
Publié: (2025)

CtrlCoT: Dual-Granularity Chain-of-Thought Compression for Controllable Reasoning
par: Fan, Zhenxuan, et autres
Publié: (2026)

OctoThinker: Mid-training Incentivizes Reinforcement Learning Scaling
par: Wang, Zengzhi, et autres
Publié: (2025)

Not All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning
par: Fu, Yu, et autres
Publié: (2024)

Beyond Token Length: Step Pruner for Efficient and Accurate Reasoning in Large Language Models
par: Wu, Canhui, et autres
Publié: (2025)

WebThinker: Empowering Large Reasoning Models with Deep Research Capability
par: Li, Xiaoxi, et autres
Publié: (2025)

RuPLaR : Efficient Latent Compression of LLM Reasoning Chains with Rule-Based Priors From Multi-Step to One-Step
par: Luo, Xiaocheng, et autres
Publié: (2026)

R3-RAG: Learning Step-by-Step Reasoning and Retrieval for LLMs via Reinforcement Learning
par: Li, Yuan, et autres
Publié: (2025)

Step-by-Step Mastery: Enhancing Soft Constraint Following Ability of Large Language Models
par: Ren, Qingyu, et autres
Publié: (2025)

Tiny Model, Big Logic: Diversity-Driven Optimization Elicits Large-Model Reasoning Ability in VibeThinker-1.5B
par: Xu, Sen, et autres
Publié: (2025)

Teaching LLMs for Step-Level Automatic Math Correction via Reinforcement Learning
par: Li, Junsong, et autres
Publié: (2025)

CoT-Valve: Length-Compressible Chain-of-Thought Tuning
par: Ma, Xinyin, et autres
Publié: (2025)

LLM Reasoners: New Evaluation, Library, and Analysis of Step-by-Step Reasoning with Large Language Models
par: Hao, Shibo, et autres
Publié: (2024)

The Zero-Step Thinking: An Empirical Study of Mode Selection as Harder Early Exit in Reasoning Models
par: Tan, Yuqiao, et autres
Publié: (2025)

Thinker: Training LLMs in Hierarchical Thinking for Deep Search via Multi-Turn Interaction
par: Xu, Jun, et autres
Publié: (2025)

Step Back to Leap Forward: Self-Backtracking for Boosting Reasoning of Language Models
par: Yang, Xiao-Wen, et autres
Publié: (2025)

R-Capsule: Compressing High-Level Plans for Efficient Large Language Model Reasoning
par: Shan, Hongyu, et autres
Publié: (2025)

Clip Your Sequences Fairly: Enforcing Length Fairness for Sequence-Level RL
par: Mao, Hanyi, et autres
Publié: (2025)

PLAN-TUNING: Post-Training Language Models to Learn Step-by-Step Planning for Complex Problem Solving
par: Parmar, Mihir, et autres
Publié: (2025)

DeepCompress: A Dual Reward Strategy for Dynamically Exploring and Compressing Reasoning Chains
par: Liang, Tian, et autres
Publié: (2025)

LAPO: Internalizing Reasoning Efficiency via Length-Adaptive Policy Optimization
par: Wu, Xingyu, et autres
Publié: (2025)

Where Does Reasoning Break? Step-Level Hallucination Detection via Hidden-State Transport Geometry
par: Alvarez, Tyler, et autres
Publié: (2026)

Multilingual LLMs Are Not Multilingual Thinkers: Evidence from Hindi Analogy Evaluation
par: Gupta, Ashray, et autres
Publié: (2025)

ParaThinker: Native Parallel Thinking as a New Paradigm to Scale LLM Test-time Compute
par: Wen, Hao, et autres
Publié: (2025)

More Thinking, More Bias: Length-Driven Position Bias in Reasoning Models
par: Wang, Xiao
Publié: (2026)