Cover Image

SafeSteer: Interpretable Safety Steering with Refusal-Evasion in LLMs

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Ghosh, Shaona, Bhattacharjee, Amrita, Ziser, Yftah, Parisien, Christopher
Format:	Preprint
Published:	2025
Subjects:	Machine Learning
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Towards Inference-time Category-wise Safety Steering for Large Language Models
by: Bhattacharjee, Amrita, et al.
Published: (2024)

AEGIS: Online Adaptive AI Content Safety Moderation with Ensemble of LLM Experts
by: Ghosh, Shaona, et al.
Published: (2024)

SafeSteer: Localized On-Policy Distillation for Efficient Safety Alignment
by: Li, Hao, et al.
Published: (2026)

Layer by Layer: Uncovering Where Multi-Task Learning Happens in Instruction-Tuned Large Language Models
by: Zhao, Zheng, et al.
Published: (2024)

Steering Language Model Refusal with Sparse Autoencoders
by: O'Brien, Kyle, et al.
Published: (2024)

AlphaSteer: Learning Refusal Steering with Principled Null-Space Constraint
by: Sheng, Leheng, et al.
Published: (2025)

Programming Refusal with Conditional Activation Steering
by: Lee, Bruce W., et al.
Published: (2024)

SteeringSafety: A Systematic Safety Evaluation Framework of Representation Steering in LLMs
by: Siu, Vincent, et al.
Published: (2025)

What Drives Representation Steering? A Mechanistic Case Study on Steering Refusal
by: Cheng, Stephen, et al.
Published: (2026)

More Test-Time Compute Can Hurt: Overestimation Bias in LLM Beam Search
by: Dalal, Gal, et al.
Published: (2026)

Neural Message-Passing on Attention Graphs for Hallucination Detection
by: Frasca, Fabrizio, et al.
Published: (2025)

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

Steering to Say No: Configurable Refusal via Activation Steering in Vision Language Models
by: Yang, Jiaxi, et al.
Published: (2026)

SteerConf: Steering LLMs for Confidence Elicitation
by: Zhou, Ziang, et al.
Published: (2025)

Beyond Token Probes: Hallucination Detection via Activation Tensors with ACT-ViT
by: Bar-Shalom, Guy, et al.
Published: (2025)

From Actions to Words: Towards Abstractive-Textual Policy Summarization in RL
by: Admoni, Sahar, et al.
Published: (2025)

BarrierSteer: LLM Safety via Learning Barrier Steering
by: Tran, Thanh Q., et al.
Published: (2026)

Feature-Guided SAE Steering for Refusal-Rate Control using Contrasting Prompts
by: Bhargav, Samaksh, et al.
Published: (2025)

HyperSteer: Activation Steering at Scale with Hypernetworks
by: Sun, Jiuding, et al.
Published: (2025)

One-shot Optimized Steering Vectors Mediate Safety-relevant Behaviors in LLMs
by: Dunefsky, Jacob, et al.
Published: (2025)

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

Learning to Steer: Input-dependent Steering for Multimodal LLMs
by: Parekh, Jayneel, et al.
Published: (2025)

Selective Safety Steering via Value-Filtered Decoding
by: Einbinder, Bat-Sheva, et al.
Published: (2026)

Surfacing Semantic Orthogonality Across Model Safety Benchmarks: A Multi-Dimensional Analysis
by: Bennion, Jonathan, et al.
Published: (2025)

Steer2Adapt: Dynamically Composing Steering Vectors Elicits Efficient Adaptation of LLMs
by: Han, Pengrui, et al.
Published: (2026)

Steering LLMs via Scalable Interactive Oversight
by: Zhou, Enyu, et al.
Published: (2026)

Steering LLMs for Formal Theorem Proving
by: Kirtania, Shashank, et al.
Published: (2025)

SafeSwitch: Steering Unsafe LLM Behavior via Internal Activation Signals
by: Han, Peixuan, et al.
Published: (2025)

The Rogue Scalpel: Activation Steering Compromises LLM Safety
by: Korznikov, Anton, et al.
Published: (2025)

Graph-Regularized Sparse Autoencoders for LLM Safety Steering
by: Yeon, Jehyeok, et al.
Published: (2025)

Interpreting and Steering Protein Language Models through Sparse Autoencoders
by: Garcia, Edith Natalia Villegas, et al.
Published: (2025)

Interpreting and Steering State-Space Models via Activation Subspace Bottlenecks
by: Mohan, Vamshi Sunku, et al.
Published: (2026)

GSS: Gated Subspace Steering for Selective Memorization Mitigation in LLMs
by: Zhang, Xuanqi, et al.
Published: (2026)

Steering Code LLMs with Activation Directions for Language and Library Control
by: Rahman, Md Mahbubur, et al.
Published: (2026)

Analyze Feature Flow to Enhance Interpretation and Steering in Language Models
by: Laptev, Daniil, et al.
Published: (2025)

Focus On This, Not That! Steering LLMs with Adaptive Feature Specification
by: Lamb, Tom A., et al.
Published: (2024)

Where to Steer: Input-Dependent Layer Selection for Steering Improves LLM Alignment
by: Gadgil, Soham, et al.
Published: (2026)

Automating Steering for Safe Multimodal Large Language Models
by: Wu, Lyucheng, et al.
Published: (2025)

Spectral Editing of Activations for Large Language Model Alignment
by: Qiu, Yifu, et al.
Published: (2024)

To Steer or Not to Steer? Mechanistic Error Reduction with Abstention for Language Models
by: Hedström, Anna, et al.
Published: (2025)