Imagen de Portada

SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Li, Hao, An, Jingkun, Song, Zijun, Zhu, Pengyu, Li, Rui, Wang, Hao, Feng, Wendi, Liu, Yesheng, Li, Lijun, Yao, Jin-Ge, Sha, Lei
Formato:	Preprint
Publicado:	2026
Materias:	Artificial Intelligence Computation and Language
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

SafeSteer: Interpretable Safety Steering with Refusal-Evasion in LLMs
por: Ghosh, Shaona, et al.
Publicado: (2025)

SafeSteer: Adaptive Subspace Steering for Efficient Jailbreak Defense in Vision-Language Models
por: Zeng, Xiyu, et al.
Publicado: (2025)

Layer-Aware Representation Filtering: Purifying Finetuning Data to Preserve LLM Safety Alignment
por: Li, Hao, et al.
Publicado: (2025)

Be Your Own Red Teamer: Safety Alignment via Self-Play and Reflective Experience Replay
por: Wang, Hao, et al.
Publicado: (2026)

SafeSteer: A Decoding-level Defense Mechanism for Multimodal Large Language Models
por: Zeng, Xinyi, et al.
Publicado: (2026)

SafeCRS: Personalized Safety Alignment for LLM-Based Conversational Recommender Systems
por: Hao, Haochang, et al.
Publicado: (2026)

Amortized Network Intervention to Steer the Excitatory Point Processes
por: Song, Zitao, et al.
Publicado: (2023)

Towards Understanding Safety Alignment: A Mechanistic Perspective from Safety Neurons
por: Chen, Jianhui, et al.
Publicado: (2024)

Pre-training Distillation for Large Language Models: A Design Space Exploration
por: Peng, Hao, et al.
Publicado: (2024)

SteerDiff: Steering towards Safe Text-to-Image Diffusion Models
por: Zhang, Hongxiang, et al.
Publicado: (2024)

Traversal-as-Policy: Log-Distilled Gated Behavior Trees as Externalized, Verifiable Policies for Safe, Robust, and Efficient Agents
por: Li, Peiran, et al.
Publicado: (2026)

Reducing the Safety Tax in LLM Safety Alignment with On-Policy Self-Distillation
por: Fu, Yu, et al.
Publicado: (2026)

SafeWatch: An Efficient Safety-Policy Following Video Guardrail Model with Transparent Explanations
por: Chen, Zhaorun, et al.
Publicado: (2024)

Multilingual Safety Alignment via Self-Distillation
por: Qin, Ruiyang, et al.
Publicado: (2026)

StepCodeReasoner: Aligning Code Reasoning with Stepwise Execution Traces via Reinforcement Learning
por: Wang, Hao, et al.
Publicado: (2026)

ASETF: A Novel Method for Jailbreak Attack on LLMs through Translate Suffix Embeddings
por: Wang, Hao, et al.
Publicado: (2024)

ROAST: Rollout-based On-distribution Activation Steering Technique
por: Su, Xuanbo, et al.
Publicado: (2026)

PsySafe: A Comprehensive Framework for Psychological-based Attack, Defense, and Evaluation of Multi-agent System Safety
por: Zhang, Zaibin, et al.
Publicado: (2024)

DAVSP: Safety Alignment for Large Vision-Language Models via Deep Aligned Visual Safety Prompt
por: Zhang, Yitong, et al.
Publicado: (2025)

DA-Mamba: Learning Domain-Aware State Space Model for Global-Local Alignment in Domain Adaptive Object Detection
por: Li, Haochen, et al.
Publicado: (2026)

UEREBot: Learning Safe Quadrupedal Locomotion under Unstructured Environments and High-Speed Dynamic Obstacles
por: Xu, Zihao, et al.
Publicado: (2026)

Effects of Low-Dose Recombinant Human Brain Natriuretic Peptide on Anterior Myocardial Infarction Complicated by Cardiogenic Shock
por: Yesheng Pan
Publicado: (2017)

Steer-MoE: Efficient Audio-Language Alignment with a Mixture-of-Experts Steering Module
por: Feng, Ruitao, et al.
Publicado: (2025)

Enhancing Peak Assignment in 13C NMR Spectroscopy: A Novel Approach Using Multimodal Alignment
por: Xu, Hao, et al.
Publicado: (2023)

Safe Transformer: An Explicit Safety Bit For Interpretable And Controllable Alignment
por: Feng, Jingyuan, et al.
Publicado: (2026)

2DNMRGym: An Annotated Experimental Dataset for Atom-Level Molecular Representation Learning in 2D NMR via Surrogate Supervision
por: Li, Yunrui, et al.
Publicado: (2025)

Harnessing the Plug-and-Play Controller by Prompting
por: Wang, Hao, et al.
Publicado: (2024)

WildReward: Learning Reward Models from In-the-Wild Human Interactions
por: Peng, Hao, et al.
Publicado: (2026)

Policy Bifurcation in Safe Reinforcement Learning
por: Zou, Wenjun, et al.
Publicado: (2024)

SafeMLRM: Demystifying Safety in Multi-modal Large Reasoning Models
por: Fang, Junfeng, et al.
Publicado: (2025)

The Extrapolation Cliff in On-Policy Distillation of Near-Deterministic Structured Outputs
por: Li, Xin, et al.
Publicado: (2026)

Engineering Epsilon‐Near‐Zero Media with Waveguides (Adv. Phys. Res. 9/2024)
por: Peihang Li, et al.
Publicado: (2024)

Robust Tensor Completion via Gradient Tensor Nulclear L1-L2 Norm for Traffic Data Recovery
por: Shu, Hao, et al.
Publicado: (2025)

Interfacial Wetting and Erosion Behavior between Fluorine‐Free Slag and Submerged Entry Nozzle under an External Electric Field
por: Zijun Peng, et al.
Publicado: (2025)

On-Policy Self-Evolution via Failure Trajectories for Agentic Safety Alignment
por: Yin, Bo, et al.
Publicado: (2026)

SafeAligner: Safety Alignment against Jailbreak Attacks via Response Disparity Guidance
por: Huang, Caishuang, et al.
Publicado: (2024)

SafeWorld: Geo-Diverse Safety Alignment
por: Yin, Da, et al.
Publicado: (2024)

SafeGRPO: Self-Rewarded Multimodal Safety Alignment via Rule-Governed Policy Optimization
por: Rong, Xuankun, et al.
Publicado: (2025)

Steering benzyl alcohol electrooxidation coupled with hydrogen evolution via hetero‐interface construction
por: Xin Du, et al.
Publicado: (2024)

STARE at the Structure: Steering ICL Exemplar Selection with Structural Alignment
por: Li, Jiaqian, et al.
Publicado: (2025)