Copertina

TokenSqueeze: Performance-Preserving Compression for Reasoning LLMs

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Zhang, Yuxiang, Yu, Zhengxu, Pan, Weihang, Jin, Zhongming, Fu, Qiang, Cai, Deng, Lin, Binbin, Ye, Jieping
Natura:	Preprint
Pubblicazione:	2025
Soggetti:	Machine Learning Artificial Intelligence
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

Enhancing Spatial Reasoning through Visual and Textual Thinking
di: Liang, Xun, et al.
Pubblicazione: (2025)

SPAR: Support-Preserving Action Rectification
di: Zhao, Jiaxin, et al.
Pubblicazione: (2026)

Enhancing Chain-of-Thought Reasoning with Critical Representation Fine-tuning
di: Huang, Chenxi, et al.
Pubblicazione: (2025)

Plan-on-Graph: Self-Correcting Adaptive Planning of Large Language Model on Knowledge Graphs
di: Chen, Liyi, et al.
Pubblicazione: (2024)

Controlling Thinking Speed in Reasoning Models
di: Lin, Zhengkai, et al.
Pubblicazione: (2025)

Token-Efficient Prompt Injection Attack: Provoking Cessation in LLM Reasoning via Adaptive Token Compression
di: Cui, Yu, et al.
Pubblicazione: (2025)

Enhancing Multiple Dimensions of Trustworthiness in LLMs via Sparse Activation Control
di: Xiao, Yuxin, et al.
Pubblicazione: (2024)

Process or Result? Manipulated Ending Tokens Can Mislead Reasoning LLMs to Ignore the Correct Reasoning Steps
di: Cui, Yu, et al.
Pubblicazione: (2025)

Meta-Soft: Leveraging Composable Meta-Tokens for Context-Preserving KV Cache Compression
di: Luo, Wei, et al.
Pubblicazione: (2026)

RL Squeezes, SFT Expands: A Comparative Study of Reasoning LLMs
di: Matsutani, Kohsei, et al.
Pubblicazione: (2025)

MemoSight: Unifying Context Compression and Multi Token Prediction for Reasoning Acceleration
di: Liu, Xinyu, et al.
Pubblicazione: (2026)

OneLatent: Single-Token Compression for Visual Latent Reasoning
di: Lv, Bo, et al.
Pubblicazione: (2026)

Pair-In, Pair-Out: Latent Multi-Token Prediction for Efficient LLMs
di: Tan, Wenhui, et al.
Pubblicazione: (2026)

TokenSkip: Controllable Chain-of-Thought Compression in LLMs
di: Xia, Heming, et al.
Pubblicazione: (2025)

When Reasoning Meets Compression: Understanding the Effects of LLMs Compression on Large Reasoning Models
di: Zhang, Nan, et al.
Pubblicazione: (2025)

Chain Of Thought Compression: A Theoretical Analysis
di: Li, Juncai, et al.
Pubblicazione: (2026)

AdaTok: Adaptive Token Compression with Object-Aware Representations for Efficient Multimodal LLMs
di: Zhang, Xinliang, et al.
Pubblicazione: (2025)

MASPO: Unifying Gradient Utilization, Probability Mass, and Signal Reliability for Robust and Sample-Efficient LLM Reasoning
di: Fu, Xiaoliang, et al.
Pubblicazione: (2026)

Through the Lens of Contrast: Self-Improving Visual Reasoning in VLMs
di: Pan, Zhiyu, et al.
Pubblicazione: (2026)

Say Anything but This: When Tokenizer Betrays Reasoning in LLMs
di: Ayoobi, Navid, et al.
Pubblicazione: (2026)

SpeContext: Enabling Efficient Long-context Reasoning with Speculative Context Sparsity in LLMs
di: Xu, Jiaming, et al.
Pubblicazione: (2025)

Model Compression and Efficient Inference for Large Language Models: A Survey
di: Wang, Wenxiao, et al.
Pubblicazione: (2024)

Not All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning
di: Fu, Yu, et al.
Pubblicazione: (2024)

One-Token Verification for Reasoning Correctness Estimation
di: Zhuang, Zhan, et al.
Pubblicazione: (2026)

A Survey of Deep Learning Methods in Protein Bioinformatics and its Impact on Protein Design
di: Dai, Weihang
Pubblicazione: (2025)

MARGE: Improving Math Reasoning for LLMs with Guided Exploration
di: Gao, Jingyue, et al.
Pubblicazione: (2025)

Don't Take Things Out of Context: Attention Intervention for Enhancing Chain-of-Thought Reasoning in Large Language Models
di: Yan, Shaotian, et al.
Pubblicazione: (2025)

CodeAgents: A Token-Efficient Framework for Codified Multi-Agent Reasoning in LLMs
di: Yang, Bruce, et al.
Pubblicazione: (2025)

Concise and Organized Perception Facilitates Reasoning in Large Language Models
di: Liu, Junjie, et al.
Pubblicazione: (2023)

Does Compression Preserve Uncertainty? A Unified Benchmark for Quantized and Sparse LLMs via Conformal Prediction
di: Tong, Yujia, et al.
Pubblicazione: (2026)

HybridToken-VLM: Hybrid Token Compression for Vision-Language Models
di: Zhang, Jusheng, et al.
Pubblicazione: (2025)

SmartThinker: Learning to Compress and Preserve Reasoning by Step-Level Length Control
di: He, Xingyang, et al.
Pubblicazione: (2025)

From Tokens to Thoughts: How LLMs and Humans Trade Compression for Meaning
di: Shani, Chen, et al.
Pubblicazione: (2025)

Prompting Multi-Modal Tokens to Enhance End-to-End Autonomous Driving Imitation Learning with LLMs
di: Duan, Yiqun, et al.
Pubblicazione: (2024)

Do LLMs Encode Functional Importance of Reasoning Tokens?
di: Singh, Janvijay, et al.
Pubblicazione: (2026)

From Skills to Talent: Organising Heterogeneous Agents as a Real-World Company
di: Yu, Zhengxu, et al.
Pubblicazione: (2026)

Tokenization Constraints in LLMs: A Study of Symbolic and Arithmetic Reasoning Limits
di: Zhang, Xiang, et al.
Pubblicazione: (2025)

Reasoning Compression with Mixed-Policy Distillation
di: Yang, Han, et al.
Pubblicazione: (2026)

Accelerating Constrained Decoding with Token Space Compression
di: Sullivan, Michael, et al.
Pubblicazione: (2026)

Thinking with Reasoning Skills: Fewer Tokens, More Accuracy
di: Zhao, Guangxiang, et al.
Pubblicazione: (2026)