Image de couverture de livre

Multilingual Safety Alignment Via Sparse Weight Editing

Fuente: arXiv

Enregistré dans:

Détails bibliographiques
Auteurs principaux:	Liang, Jiaming, Wang, Zhaoxin, Wang, Handing
Format:	Preprint
Publié:	2026
Sujets:	Machine Learning
Accès en ligne:	Acceder al recurso
Tags:	Ajouter un tag Pas de tags, Soyez le premier à ajouter un tag!

Documents similaires

SafeNeuron: Neuron-Level Safety Alignment for Large Language Models
par: Wang, Zhaoxin, et autres
Publié: (2026)

Implicit Jailbreak Attacks via Cross-Modal Information Concealment on Vision-Language Models
par: Wang, Zhaoxin, et autres
Publié: (2025)

Enhancing the Effectiveness and Durability of Backdoor Attacks in Federated Learning through Maximizing Task Distinction
par: Wang, Zhaoxin, et autres
Publié: (2025)

Preventing Catastrophic Overfitting in Fast Adversarial Training: A Bi-level Optimization Perspective
par: Wang, Zhaoxin, et autres
Publié: (2024)

Multilingual Safety Alignment via Self-Distillation
par: Qin, Ruiyang, et autres
Publié: (2026)

Safety-Critical Scenario Generation Via Reinforcement Learning Based Editing
par: Liu, Haolan, et autres
Publié: (2023)

ParetoHqD: Fast Offline Multiobjective Alignment of Large Language Models using Pareto High-quality Data
par: Gu, Haoran, et autres
Publié: (2025)

Align Once, Benefit Multilingually: Enforcing Multilingual Consistency for LLM Safety Alignment
par: Bu, Yuyan, et autres
Publié: (2026)

RASA: Routing-Aware Safety Alignment for Mixture-of-Experts Models
par: Liang, Jiacheng, et autres
Publié: (2026)

Training In-Context and In-Weights Mixtures Via Contrastive Context Sampling
par: Malu, Deeptanshu, et autres
Publié: (2026)

DTVI: Dual-Stage Textual and Visual Intervention for Safe Text-to-Image Generation
par: Tan, Binhong, et autres
Publié: (2026)

EASE: Practical and Efficient Safety Alignment for Small Language Models
par: Shi, Haonan, et autres
Publié: (2025)

Few Tokens, Big Leverage: Preserving Safety Alignment by Constraining Safety Tokens during Fine-tuning
par: Wang, Guoli, et autres
Publié: (2026)

Model-Editing-Based Jailbreak against Safety-aligned Large Language Models
par: Li, Yuxi, et autres
Publié: (2024)

From Parameter to Representation: A Closed-Form Approach for Controllable Model Merging
par: Wu, Jialin, et autres
Publié: (2025)

Controllable Value Alignment in Large Language Models through Neuron-Level Editing
par: Yang, Yonghui, et autres
Publié: (2026)

Equivariant Deep Weight Space Alignment
par: Navon, Aviv, et autres
Publié: (2023)

Variance Reduction and Low Sample Complexity in Stochastic Optimization via Proximal Point Method
par: Liang, Jiaming
Publié: (2024)

Unlocking the Address Book: Dissecting the Sparse Semantic Structure of LLM Key-Value Caches via Sparse Autoencoders
par: Ma, Qingsen, et autres
Publié: (2025)

Rethinking Graph Masked Autoencoders through Alignment and Uniformity
par: Wang, Liang, et autres
Publié: (2024)

Stable and Steerable Sparse Autoencoders with Weight Regularization
par: Jedryszek, Piotr, et autres
Publié: (2026)

SaLoRA: Safety-Alignment Preserved Low-Rank Adaptation
par: Li, Mingjie, et autres
Publié: (2025)

SAEMark: Steering Personalized Multilingual LLM Watermarks with Sparse Autoencoders
par: Yu, Zhuohao, et autres
Publié: (2025)

Advancing LLM Safe Alignment with Safety Representation Ranking
par: Du, Tianqi, et autres
Publié: (2025)

Safety Alignment as Continual Learning: Mitigating the Alignment Tax via Orthogonal Gradient Projection
par: Sun, Guanglong, et autres
Publié: (2026)

SparsePO: Controlling Preference Alignment of LLMs via Sparse Token Masks
par: Christopoulou, Fenia, et autres
Publié: (2024)

Reducing the Safety Tax in LLM Safety Alignment with On-Policy Self-Distillation
par: Fu, Yu, et autres
Publié: (2026)

AlphaAlign: Incentivizing Safety Alignment with Extremely Simplified Reinforcement Learning
par: Zhang, Yi, et autres
Publié: (2025)

Stream Aligner: Efficient Sentence-Level Alignment via Distribution Induction
par: Lou, Hantao, et autres
Publié: (2025)

Lifelong Safety Alignment for Language Models
par: Wang, Haoyu, et autres
Publié: (2025)

Soft Quantization: Model Compression Via Weight Coupling
par: Bernstein, Daniel T., et autres
Publié: (2026)

Hierarchical Structure-Property Alignment for Data-Efficient Molecular Generation and Editing
par: Fan, Ziyu, et autres
Publié: (2025)

Edit Less, Achieve More: Dynamic Sparse Neuron Masking for Lifelong Knowledge Editing in LLMs
par: Liu, Jinzhe, et autres
Publié: (2025)

Curriculum Learning for Safety Alignment
par: Kumar, Sandeep, et autres
Publié: (2026)

LLM Safety Alignment is Divergence Estimation in Disguise
par: Haldar, Rajdeep, et autres
Publié: (2025)

EvoEdit: Evolving Null-space Alignment for Robust and Efficient Knowledge Editing
par: Lyu, Sicheng, et autres
Publié: (2025)

Hierarchical Multiple Kernel K-Means Algorithm Based on Sparse Connectivity
par: Wang, Lei, et autres
Publié: (2024)

Enhance the Safety in Reinforcement Learning by ADRC Lagrangian Methods
par: Zhang, Mingxu, et autres
Publié: (2026)

Towards Understanding Safety Alignment: A Mechanistic Perspective from Safety Neurons
par: Chen, Jianhui, et autres
Publié: (2024)

Safety-Utility Conflicts Are Not Global: Surgical Alignment via Head-Level Diagnosis
par: Cai, Wang, et autres
Publié: (2026)