Copertina

Meaningless Tokens, Meaningful Gains: How Activation Shifts Enhance LLM Reasoning

Fuente: arXiv

Salvato in:

Dettagli Bibliografici
Autori principali:	Shi, Zeru, Wan, Yingjia, Wang, Zhenting, Wang, Qifan, Yang, Fan, Kreiss, Elisa, Tang, Ruixiang
Natura:	Preprint
Pubblicazione:	2025
Soggetti:	Machine Learning
Accesso online:	Acceder al recurso
Tags:	Aggiungi Tag Nessun Tag, puoi essere il primo ad aggiungerne!!

Documenti analoghi

A Single Layer to Explain Them All:Understanding Massive Activations in Large Language Models
di: Shi, Zeru, et al.
Pubblicazione: (2026)

Read the Scene, Not the Script: Outcome-Aware Safety for LLMs
di: Wu, Rui, et al.
Pubblicazione: (2025)

Auto-Prompt Generation is Not Robust: Prompt Optimization Driven by Pseudo Gradient
di: Shi, Zeru, et al.
Pubblicazione: (2024)

Token-Budget-Aware LLM Reasoning
di: Han, Tingxu, et al.
Pubblicazione: (2024)

Micro-Defects Expose Macro-Fakes: Detecting AI-Generated Images via Local Distributional Shifts
di: Zhang, Boxuan, et al.
Pubblicazione: (2026)

Enhanced Graph Transformer with Serialized Graph Tokens
di: Wang, Ruixiang, et al.
Pubblicazione: (2026)

Reasoning Shift: How Context Silently Shortens LLM Reasoning
di: Rodionov, Gleb
Pubblicazione: (2026)

Pruning the Unsurprising: Efficient LLM Reasoning via First-Token Surprisal
di: Zeng, Wenhao, et al.
Pubblicazione: (2025)

Critical Tokens Matter: Token-Level Contrastive Estimation Enhances LLM's Reasoning Capability
di: Lin, Zicheng, et al.
Pubblicazione: (2024)

Depth-Breadth Synergy in RLVR: Unlocking LLM Reasoning Gains with Adaptive Exploration
di: Yang, Zhicheng, et al.
Pubblicazione: (2025)

DUMP: Automated Distribution-Level Curriculum Learning for RL-based LLM Post-training
di: Wang, Zhenting, et al.
Pubblicazione: (2025)

Shifting Uncertainty to Critical Moments: Towards Reliable Uncertainty Quantification for VLA Model
di: Tang, Yanchuan, et al.
Pubblicazione: (2026)

Time-Shifted Token Scheduling for Symbolic Music Generation
di: Wang, Ting-Kang, et al.
Pubblicazione: (2025)

Memex(RL): Scaling Long-Horizon LLM Agents via Indexed Experience Memory
di: Wang, Zhenting, et al.
Pubblicazione: (2026)

ADO: Automatic Data Optimization for Inputs in LLM Prompts
di: Lin, Sam, et al.
Pubblicazione: (2025)

Beyond Differences: Doubly Robust Meta-Learners for Ratio-Based Treatment Effects
di: Fuchs, Michael, et al.
Pubblicazione: (2026)

ProLLM: Protein Chain-of-Thoughts Enhanced LLM for Protein-Protein Interaction Prediction
di: Jin, Mingyu, et al.
Pubblicazione: (2024)

Token-Efficient RL for LLM Reasoning
di: Lee, Alan, et al.
Pubblicazione: (2025)

Not All Tokens Matter: Towards Efficient LLM Reasoning via Token Significance in Reinforcement Learning
di: Liu, Hanbing, et al.
Pubblicazione: (2025)

Enhanced Federated Optimization: Adaptive Unbiased Client Sampling with Reduced Variance
di: Zeng, Dun, et al.
Pubblicazione: (2023)

BiGain: Unified Token Compression for Joint Generation and Classification
di: Liu, Jiacheng, et al.
Pubblicazione: (2026)

Testing for correlation between network structure and high-dimensional node covariates
di: Fuchs-Kreiss, Alexander, et al.
Pubblicazione: (2025)

Beyond Reasoning Gains: Mitigating General Capabilities Forgetting in Large Reasoning Models
di: Phan, Hoang, et al.
Pubblicazione: (2025)

When Reward Hacking Rebounds: Understanding and Mitigating It with Representation-Level Signals
di: Wu, Rui, et al.
Pubblicazione: (2026)

Gained in Translation: Privileged Pairwise Judges Enhance Multilingual Reasoning
di: Sutawika, Lintang, et al.
Pubblicazione: (2026)

Enhancing Low-Precision Sampling via Stochastic Gradient Hamiltonian Monte Carlo
di: Wang, Ziyi, et al.
Pubblicazione: (2023)

Visual Agents as Fast and Slow Thinkers
di: Sun, Guangyan, et al.
Pubblicazione: (2024)

Accordion-Thinking: Self-Regulated Step Summaries for Efficient and Readable LLM Reasoning
di: Yang, Zhicheng, et al.
Pubblicazione: (2026)

TokenSeek: Memory Efficient Fine Tuning via Instance-Aware Token Ditching
di: Zeng, Runjia, et al.
Pubblicazione: (2026)

The Reasoning Trap: How Enhancing LLM Reasoning Amplifies Tool Hallucination
di: Yin, Chenlong, et al.
Pubblicazione: (2025)

Understanding the Effect of using Semantically Meaningful Tokens for Visual Representation Learning
di: Kalibhat, Neha, et al.
Pubblicazione: (2024)

BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokens
di: Wen, Hao, et al.
Pubblicazione: (2025)

K2-V2: A 360-Open, Reasoning-Enhanced LLM
di: K2 Team, et al.
Pubblicazione: (2025)

A Single Revision Step Improves Token-Efficient LLM Reasoning
di: Zhang, Yingchuan, et al.
Pubblicazione: (2026)

Navigate the Unknown: Enhancing LLM Reasoning with Intrinsic Motivation Guided Exploration
di: Gao, Jingtong, et al.
Pubblicazione: (2025)

EPO: Entropy-regularized Policy Optimization for LLM Agents Reinforcement Learning
di: Xu, Wujiang, et al.
Pubblicazione: (2025)

Guiding Language Model Reasoning with Planning Tokens
di: Wang, Xinyi, et al.
Pubblicazione: (2023)

Policy Guided Tree Search for Enhanced LLM Reasoning
di: Li, Yang
Pubblicazione: (2025)

Re-Imagining Multimodal Instruction Tuning: A Representation View
di: Liu, Yiyang, et al.
Pubblicazione: (2025)

How Much Thinking is Enough? Quantifying and Understanding Redundancy in LLM Reasoning
di: Zhai, Zhiyuan, et al.
Pubblicazione: (2026)