VLM Interpretability

VLM: Vision-Language Model

Follow this topic Ask about this topic

Momentum

24 papers in the last four weeks, up 200% on the four weeks before. 0.2% of all new papers.

Jul 13Week of Sep 28

Latest papers 158

All topics

Mechanistic Insights into Functional Sparsity in Multimodal LLMs via CoRe Heads
Jun 4, 2026Ruoxi Sun, Quantong Qiu, Juntao Li +3Multimodal Large Language Models VLM Interpretability
Soochow University · Peking University
Visual Instruction Tuning Aligns Modalities through Abstraction
Jun 2, 2026Luis Palacios, Lorenzo Basile, Diego Doimo +1VLM Adaptation VLM Interpretability
Area Science Park, Trieste, Italy
Beyond Compression: Quantifying Spectral Accessibility in Vision Representations
Jun 2, 2026Akayou A. Kitessa, Yijun ZhaoVisual Representation Learning Frequency-Domain Feature Learning
Fordham University New York City, NY
Decomposing how prompting steers behavior
Jun 2, 2026Fan L. Cheng, Nikolaus KriegeskorteLLM Prompting LLM Interpretability
Columbia University
Encoded but Not Routed: Explaining the Table-Chart Gap in Scientific Claim Verification
Jun 1, 2026Sunisth Kumar, Xanh Ho, Tim Schopf +3VLM Interpretability Chart Understanding
The University of Tokyo · NII LLMC · National Institute of Informatics +2
Self-Improving Small Object Grounding in LVLMs
Jun 1, 2026Tianze Yang, Yucheng Shi, Ruitong Sun +2Large Vision-Language Models VLM Interpretability
University of Georgia
Beyond Visual Memory: Mechanistic Diagnostics of Latent Visual Reasoning
May 31, 2026Garvin Guo, Yu Chen, Xiang Wang +4VLM Evaluation VLM Interpretability
Amap, Alibaba Group · Shanghai Innovation Institute
MLLM-Microscope: Unlocking Hidden Structure Within Multimodal Large Language Models
May 30, 2026Ravil Mussabayev, Rustam MussabayevRepresentation Geometry Multimodal Large Language Models
Satbayev University
Through the PRISM: Principle-Aware, Interpretable, and Multi-Scale Evaluation of Visual Designs
May 30, 2026Mona Gandhi, KJ Joseph, Srinivasan Parthasarathy +1VLM Evaluation VLM Interpretability
Ohio State University · Adobe Research
Vision-Language Models Suppress Female Representations Under Ambiguous Input
May 29, 2026Arnau Marin-Llobet, Simon Henniger, Mahzarin R. BanajiVLM Evaluation VLM Interpretability
School of Engineering and Applied Sciences · Department of Psychology
The Regularizing Power of Language-Training Deepfake Detectors
May 29, 2026Benedikt Hopf, Zongwei Wu, Radu TimofteVLM Robustness Explainable Deepfake Detection
Computer Vision Lab, CAIDAS, University of Würzburg, Germany
Why Far Looks Up: Probing Spatial Representation in Vision-Language Models
May 28, 2026Cheolhong Min, Jaeyun Jung, Daeun Lee +5VLM Evaluation Spatial Reasoning Benchmarks
Seoul National University · The Ohio State University · NVIDIA
VITAL: Visual-Semantic Dual Supervision for Enhanced and Interpretable Latent Reasoning in Medical MLLMs
May 27, 2026Qiaoru Li, Shaotian Liang, Jintao Chen +4Medical VQA VLM Interpretability
3Tencent · 1Zhejiang University · 2Shanghai AI Laboratory +1
Explaining is Harder Than Predicting Alone: Evaluating Concept-based Explanations of MLLMs as ICL Visual Classifiers
May 27, 2026Carmen Quiles-Ramírez, Leticia L. Rodríguez, Nicolás Martorell +1VLM Evaluation Multimodal Large Language Models
Universidad de Granada, Spain · Universidad de Buenos Aires, Argentina
Real Images, Worse Judgments: Evaluating Vision-Language Models on Concreteness and Imagery
May 26, 2026Yifan Jiang, Ruoxi Ning, Sheng Yao +1VLM Evaluation Vision-Language Models
University of Waterloo · Vector Institute
Can Retrieval Heads See Images? Multimodal Retrieval Heads in Long-Context Vision-Language Models
May 26, 2026Aaron Branson Cigres Li, Zhaowei Wang, Yu Zhao +9Vision-Language Models VLM Interpretability
1HKUST · University of Edinburgh · 4CUHK +2
Binding Visual Features Point by Point
May 25, 2026Udith Haputhanthri, Declan Campbell, Rim Assouel +2Vision-Language Models VLM Interpretability
Princeton University · Mila – Quebec AI Institute · Université de Montréal
Interpretability Transfer from Language to Vision via Sparse Autoencoders
May 24, 2026Alexey Kravets, Da Li, Chuan Li +2VLM Interpretability Sparse Autoencoders
Seeing without Looking: Do Vision-Language Benchmarks Really Test Vision?
May 21, 2026Zixuan Lan, Luzhe Sun, Matthew R. Walter +1VLM Evaluation VLM Interpretability
University of Chicago · Toyota Technological Institute at Chicago · Stony Brook University
Transcoders Trace Visual Grounding and Hallucinations in Vision-Language Models
May 21, 2026Dimitrios Damianos, Leon Voukoutis, Georgios Skyrianos +2Object Hallucination in VLMs VLM Interpretability
Institute of Language and Speech Processing, Athena Research Center Athens, Greece
Conceptualizing Embeddings: Sparse Disentanglement for Vision-Language Models
May 21, 2026Piotr Kubaty, Patryk Marszałek, Łukasz Struski +3Disentangled Representation Learning Multimodal Disentangled Representation Learning
Faculty of Mathematics and Computer Science, Jagiellonian University, Kraków, Poland · Doctoral School of Exact and Natural Sciences, Jagiellonian University, Kraków, Poland
SegCompass: Exploring Interpretable Alignment with Sparse Autoencoders for Enhanced Reasoning Segmentation
May 21, 2026Zhenyu Lu, Liupeng Li, Jinpeng Wang +4Image Segmentation Text-Guided Image Segmentation
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences · Peng Cheng Laboratory · University of Chinese Academy of Sciences +3
Measuring Cross-Modal Synergy: A Benchmark for VLM Explainability
May 21, 2026Joël Roman Ky, Salah Ghamizi, Maxime CordyVLM Interpretability Shapley Value Attribution
University of Luxembourg, Luxembourg · Luxembourg Institute of Health (LIH), Luxembourg
Interpreting and Enhancing Emotional Circuits in Large Vision-Language Models via Cross-Modal Information Flow
May 21, 2026Chengsheng Zhang, Chenghao Sun, Zhining Xie +1VLM Hallucination VLM Interpretability
MoE Key Laboratory of Brain-inspired Intelligent Perception and Cognition, University of Science and Technology of China · AIPD, Tencent
Ablate-to-Validate: Are Vision-Language Models Really Using Continuous Thought Tokens?
May 20, 2026Tianyi Zhang, Mahtab Bigverdi, Ranjay KrishnaVLM Evaluation VLM Interpretability
University of Washington
Do Vision-Language Models Understand 3D Scenes or Just Catalogue Objects?
May 19, 2026Animesh Maheshwari, Divyansh Sahu, Nishit VermaVLM Evaluation Spatial Reasoning Benchmarks
Deccan AI
Rethinking Visual Attribution for Chest X-ray Reasoning in Large Vision Language Models
May 19, 2026Guangzhi Xiong, Qiao Jin, Sanchit Sinha +2Medical VQA VLM Interpretability
University of Virginia · National Institutes of Health
Mechanisms of Object Localization in Vision-Language Models
May 19, 2026Timothy Schaumlöffel, Martina G. Vilas, Gemma RoigVision-Language Models VLM Interpretability
Goethe University Frankfurt, Germany · The Hessian Center for AI, Germany
Causal Evidence for Attention Head Imbalance in Modality Conflict Hallucination
May 19, 2026Jinrui Jiang, Zhangtai Wu, Zhen Wu +1Hallucination in Language Models Multimodal Large Language Models
National Key Laboratory for Novel Software Technology, Nanjing University, China · School of Artificial Intelligence, Nanjing University, China
Employing Vision-Language Models for Face Image Quality Assessment
May 17, 2026Erdi Sarıtaş, Eren Onaran, Vitomir Štruc +1VLM Evaluation VLM Robustness
Department of Computer Engineering, Istanbul Technical University, Istanbul, Türkiye · Faculty of Electrical Engineering, University of Ljubljana, Ljubljana, Slovenia · Division of Engineering, NYU Abu Dhabi, Abu Dhabi, UAE

Jun 4, 2026, cs.CLJ/K move · Enter open · S save

Mechanistic Insights into Functional Sparsity in Multimodal LLMs via CoRe Heads

Ruoxi Sun, Quantong Qiu, Juntao Li, Zecheng Tang+2

Soochow University · Peking University

Multimodal Large Language Models VLM Interpretability Attention Head Specialization

Open paper AskPDF arXiv

While Multimodal Large Language Models (MLLMs) demonstrate remarkable proficiency on complex vision-language tasks, the mechanisms by which they extract query-relevant visual features from complex, noisy contexts remain opaque. In this paper, we present an in-depth interpretability study that uncovers a profound structural property within MLLMs: functional sparsity in cross-modal retrieval. Leveraging a token-level metric termed Retrieval Attention Mass (RAM), we identify and characterize a highly specialized subset of attention heads, referred to as Context-aware Retrieval (CoRe) heads. Across diverse visual domains and model scales, we observe a clear functional division: CoRe heads act as dedicated information extractors, while most other heads distribute attention over broader contextual regions. Causal interventions further demonstrate the necessity of these specialized heads. Ablating only the top 5% of CoRe heads causes significant degradation in multimodal reasoning performance, whereas ablating lower-ranked heads has minimal effect. Moreover, acceleration experiments validate the utility of CoRe heads, showing that leveraging this localized sparsity significantly accelerates inference while maintaining robust task performance. Our findings reveal a structural principle of functional sparsity within MLLMs, refining the current understanding of mechanistic interpretability and laying a theoretical foundation that can inspire future architecture design and model optimization.

Momentum

Related topics

Latest papers 158