Cover Image

Refusal Direction is Universal Across Safety-Aligned Languages

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Wang, Xinpeng, Wang, Mingyang, Liu, Yihong, Schütze, Hinrich, Plank, Barbara
Format:	Preprint
Published:	2025
Subjects:	Computation and Language
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Think Before Refusal : Triggering Safety Reflection in LLMs to Mitigate False Refusal Behavior
by: Si, Shengyun, et al.
Published: (2025)

On the Entity-Level Alignment in Crosslingual Consistency
by: Liu, Yihong, et al.
Published: (2025)

OFA: A Framework of Initializing Unseen Subword Embeddings for Efficient Large-scale Multilingual Continued Pretraining
by: Liu, Yihong, et al.
Published: (2023)

LangSAMP: Language-Script Aware Multilingual Pretraining
by: Liu, Yihong, et al.
Published: (2024)

Surgical, Cheap, and Flexible: Mitigating False Refusal in Language Models via Single Vector Ablation
by: Wang, Xinpeng, et al.
Published: (2024)

Evaluating Contextually Mediated Factual Recall in Multilingual Large Language Models
by: Liu, Yihong, et al.
Published: (2026)

A Comprehensive Evaluation of Multilingual Chain-of-Thought Reasoning: Performance, Consistency, and Faithfulness Across Languages
by: Zhao, Raoyuan, et al.
Published: (2025)

Tracing Multilingual Factual Knowledge Acquisition in Pretraining
by: Liu, Yihong, et al.
Published: (2025)

What Do Dialect Speakers Want? A Survey of Attitudes Towards Language Technology for German Dialects
by: Blaschke, Verena, et al.
Published: (2024)

MaiBaam: A Multi-Dialectal Bavarian Universal Dependency Treebank
by: Blaschke, Verena, et al.
Published: (2024)

Breaking the Script Barrier in Multilingual Pre-Trained Language Models with Transliteration-Based Post-Training Alignment
by: Xhelili, Orgest, et al.
Published: (2024)

The Anatomy of an Edit: Mechanism-Guided Activation Steering for Knowledge Editing
by: Cao, Yuan, et al.
Published: (2026)

Why Better Cross-Lingual Alignment Fails for Better Cross-Lingual Transfer: Case of Encoders
by: Veitsman, Yana, et al.
Published: (2026)

Lost in Multilinguality: Dissecting Cross-lingual Factual Inconsistency in Transformer Language Models
by: Wang, Mingyang, et al.
Published: (2025)

HYPEROFA: Expanding LLM Vocabulary to New Languages via Hypernetwork-Based Embedding Initialization
by: Özeren, Enes, et al.
Published: (2025)

BlackboxNLP-2025 MIB Shared Task: Exploring Ensemble Strategies for Circuit Localization Methods
by: Mondorf, Philipp, et al.
Published: (2025)

TransliCo: A Contrastive Learning Framework to Address the Script Barrier in Multilingual Pretrained Language Models
by: Liu, Yihong, et al.
Published: (2024)

TransMI: A Framework to Create Strong Baselines from Multilingual Pretrained Language Models for Transliterated Data
by: Liu, Yihong, et al.
Published: (2024)

Relational Linearity is a Predictor of Hallucinations
by: Lu, Yuetian, et al.
Published: (2026)

How Programming Concepts and Neurons Are Shared in Code Language Models
by: Kargaran, Amir Hossein, et al.
Published: (2025)

Your Pretrained Model Tells the Difficulty Itself: A Self-Adaptive Curriculum Learning Paradigm for Natural Language Understanding
by: Feng, Qi, et al.
Published: (2025)

Exploring the Role of Transliteration in In-Context Learning for Low-resource Languages Written in Non-Latin Scripts
by: Ma, Chunlan, et al.
Published: (2024)

Left, Right, or Center? Evaluating LLM Framing in News Classification and Generation
by: Kennedy, Molly, et al.
Published: (2026)

Understanding In-Context Machine Translation for Low-Resource Languages: A Case Study on Manchu
by: Pei, Renhao, et al.
Published: (2025)

NLNDE at SemEval-2023 Task 12: Adaptive Pretraining and Source Language Selection for Low-Resource Multilingual Sentiment Analysis
by: Wang, Mingyang, et al.
Published: (2023)

Rehearsal-Free Modular and Compositional Continual Learning for Language Models
by: Wang, Mingyang, et al.
Published: (2024)

Evaluating Robustness of Large Language Models Against Multilingual Typographical Errors
by: Zhao, Raoyuan, et al.
Published: (2025)

Language Mixing in Reasoning Language Models: Patterns, Impact, and Internal Causes
by: Wang, Mingyang, et al.
Published: (2025)

MoSECroT: Model Stitching with Static Word Embeddings for Crosslingual Zero-shot Transfer
by: Ye, Haotian, et al.
Published: (2024)

Better Call SAUL: Fluent and Consistent Language Model Editing with Generation Regularization
by: Wang, Mingyang, et al.
Published: (2024)

Large Reasoning Models Are (Not Yet) Multilingual Latent Reasoners
by: Liu, Yihong, et al.
Published: (2026)

Crosslingual On-Policy Self-Distillation for Multilingual Reasoning
by: Liu, Yihong, et al.
Published: (2026)

Enhancing Robustness of Autoregressive Language Models against Orthographic Attacks via Pixel-based Approach
by: Yang, Han, et al.
Published: (2025)

On Relation-Specific Neurons in Large Language Models
by: Liu, Yihong, et al.
Published: (2025)

Calibration Is Not Enough: Evaluating Confidence Estimation Under Language Variations
by: Xia, Yuxi, et al.
Published: (2026)

GLUScope: A Tool for Analyzing GLU Neurons in Transformer Language Models
by: Gerstner, Sebastian, et al.
Published: (2026)

DiM\textsuperscript{3}: Bridging Multilingual and Multimodal Models via Direction- and Magnitude-Aware Merging
by: Wang, Zijing, et al.
Published: (2026)

Mechanistic Understanding and Mitigation of Language Confusion in English-Centric Large Language Models
by: Nie, Ercong, et al.
Published: (2025)

ReverseMath: Answer Inversion for Scalable and Verifiable Mathematical Problem Generation
by: Zhao, Raoyuan, et al.
Published: (2026)

Learn it or Leave it: Module Composition and Pruning for Continual Learning
by: Wang, Mingyang, et al.
Published: (2024)