MAST: Medoid-Anchored Subspace Steering for Visual Grounding in Multimodal LLMs
cs.CV, cs.LG
Submitted: 2026-11-11
Updated: 2026-11-11
Terminology
Sources
- Improving Steering Vectors by Targeting Sparse Autoencoder Features
- Textual Steering Vectors Can Improve Visual Understanding in Multimodal Large Language Models
- AMPS: Adaptive Modality Preference Steering via Functional Entropy
- T-VSS: Test-Time Visual Subspace Steering for Adversarial Robustness of Vision-Language Models
- Enhancing Instruction Following of LLMs via Activation Steering with Dynamic Rejection
- Prototype-Based Dynamic Steering for Large Language Models
- Suppressing Prior-Comparison Hallucinations in Radiology Report Generation via Semantically Decoupled Latent Steering
- UMAP: Uniform Manifold Approximation and Projection for Dimension Reduction
- VISOR: Visual Input-based Steering for Output Redirection in Vision-Language Models
- Steering Language Models With Activation Engineering
- Mitigating Content Effects on Reasoning in Language Models through Fine-Grained Activation Steering
- Differentiated Directional Intervention A Framework for Evading LLM Safety Alignment
- Representation Engineering: A Top-Down Approach to AI Transparency
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs