Imagen de Portada

Steering Language Models Before They Speak: Logit-Level Interventions

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	An, Hyeseon, Park, Shinwoo, Jin, Hyundong, Han, Yo-Sub
Formato:	Preprint
Publicado:	2026
Materias:	Computation and Language Artificial Intelligence
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

DLM-SWAI: Steering Diffusion Language Models Before They Unmask
por: An, Hyeseon, et al.
Publicado: (2026)

Linguistics-Aware Non-Distortionary LLM Watermarking
por: Park, Shinwoo, et al.
Publicado: (2026)

A Linguistics-Aware LLM Watermarking via Syntactic Predictability
por: Park, Shinwoo, et al.
Publicado: (2025)

EPIC: Efficient and Parallel Inference under CFG Constraints for Diffusion Language Models
por: Jin, Hyundong, et al.
Publicado: (2026)

From Intuition to Calibrated Judgment: A Rubric-Based Expert-Panel Study of Human Detection of LLM-Generated Korean Text
por: Park, Shinwoo, et al.
Publicado: (2026)

NCO: A Versatile Plug-in for Handling Negative Constraints in Decoding
por: Jin, Hyundong, et al.
Publicado: (2026)

Sequential Behavioral Watermarking for LLM Agents
por: An, Hyeseon, et al.
Publicado: (2026)

DITTO: A Spoofing Attack Framework on Watermarked LLMs via Knowledge Distillation
por: An, Hyeseon, et al.
Publicado: (2025)

WaterMod: Modular Token-Rank Partitioning for Probability-Balanced LLM Watermarking
por: Park, Shinwoo, et al.
Publicado: (2025)

KatFishNet: Detecting LLM-Generated Korean Text through Linguistic Feature Analysis
por: Park, Shinwoo, et al.
Publicado: (2025)

RV-HATE: Reinforced Multi-Module Voting for Implicit Hate Speech Detection
por: Lee, Yejin, et al.
Publicado: (2025)

TRAPDOC: Deceiving LLM Users by Injecting Imperceptible Phantom Tokens into Documents
por: Jin, Hyundong, et al.
Publicado: (2025)

Detection of LLM-Paraphrased Code and Identification of the Responsible LLM Using Coding Style Features
por: Park, Shinwoo, et al.
Publicado: (2025)

RegexPSPACE: A Benchmark for Evaluating LLM Reasoning on PSPACE-complete Regex Problems
por: Jin, Hyundong, et al.
Publicado: (2025)

AmpleHate: Amplifying the Attention for Versatile Implicit Hate Detection
por: Lee, Yejin, et al.
Publicado: (2025)

Marking Code Without Breaking It: Code Watermarking for Detecting LLM-Generated Code
por: Kim, Jungin, et al.
Publicado: (2025)

Obfuscation Rules for Detecting and Detoxifying Korean Toxicity
por: Lee, Yejin, et al.
Publicado: (2025)

Adaptive Steering and Remasking for Safe Generation in Diffusion Language Models
por: Lee, Yejin, et al.
Publicado: (2026)

Thinking Before Speaking: A Role-playing Model with Mindset
por: Zhang, Baohua, et al.
Publicado: (2024)

How Does the Thinking Step Influence Model Safety? An Entropy-based Safety Reminder for LRMs
por: Kim, Su-Hyeon, et al.
Publicado: (2026)

The Information Geometry of Softmax: Probing and Steering
por: Park, Kiho, et al.
Publicado: (2026)

CogSteer: Cognition-Inspired Selective Layer Intervention for Efficiently Steering Large Language Models
por: Wang, Xintong, et al.
Publicado: (2024)

Think Before You Speak: Cultivating Communication Skills of Large Language Models via Inner Monologue
por: Zhou, Junkai, et al.
Publicado: (2023)

CRaFT: Circuit-Guided Refusal Feature Selection via Cross-Layer Transcoders
por: Kim, Su-Hyeon, et al.
Publicado: (2026)

Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking
por: Zelikman, Eric, et al.
Publicado: (2024)

STAB: Specification-driven Testing for Algorithmic Bottlenecks
por: Lim, Soohan, et al.
Publicado: (2026)

Prompt-Activation Duality: Improving Activation Steering via Attention-Level Interventions
por: Kang, Diancheng, et al.
Publicado: (2026)

Multi-Attribute Steering of Language Models via Targeted Intervention
por: Nguyen, Duy, et al.
Publicado: (2025)

The Linear Representation Hypothesis and the Geometry of Large Language Models
por: Park, Kiho, et al.
Publicado: (2023)

Spectral Logit Sculpting: Adaptive Low-Rank Logit Transformation for Controlled Text Generation
por: Li, Jin, et al.
Publicado: (2025)

SpeakRL: Synergizing Reasoning, Speaking, and Acting in Language Models with Reinforcement Learning
por: Acikgoz, Emre Can, et al.
Publicado: (2025)

Self-Steering Language Models
por: Grand, Gabriel, et al.
Publicado: (2025)

Steering Without Breaking: Mechanistically Informed Interventions for Discrete Diffusion Language Models
por: Zhou, Hanhan, et al.
Publicado: (2026)

Steering When Necessary: Flexible Steering Large Language Models with Backtracking
por: Cheng, Zifeng, et al.
Publicado: (2025)

Towards Reliable Evaluation of Behavior Steering Interventions in LLMs
por: Pres, Itamar, et al.
Publicado: (2024)

The Geometry of Categorical and Hierarchical Concepts in Large Language Models
por: Park, Kiho, et al.
Publicado: (2024)

Cite Before You Speak: Enhancing Context-Response Grounding in E-commerce Conversational LLM-Agents
por: Zeng, Jingying, et al.
Publicado: (2025)

UNDIAL: Self-Distillation with Adjusted Logits for Robust Unlearning in Large Language Models
por: Dong, Yijiang River, et al.
Publicado: (2024)

On the Limitations of Steering in Language Model Alignment
por: Niranjan, Chebrolu, et al.
Publicado: (2025)

Word Embeddings Are Steers for Language Models
por: Han, Chi, et al.
Publicado: (2023)