Multimodal Model Interpretability

Overview

It asks which modalities, regions, tokens, concepts, or internal circuits support a model’s output, helping people audit errors, shortcuts, bias, and safety. Common approaches include saliency and Shapley attribution, concept bottlenecks and sparse autoencoders, and probing or causal interventions that test what happens when evidence is altered. Researchers evaluate these explanations on visual question answering, hallucination and grounding tasks, cross-modal benchmarks, and clinical settings where inputs may be missing or incomplete. A central tension is that an explanation can look plausible without being faithful: models may encode useful evidence yet ignore it, favor one modality, or rely on shortcuts, and evaluation of genuine cross-modal reasoning remains difficult.

Latest papers

CardsList

141 papers15 papers in the last four weeks, up 36% on the four weeks before. 0.1% of all new papers.

  1. The Elephant in the Question: Unified Models Picture the Concepts Behind Their Answers

    Nov 18, 2026Tobias Braun, Marcus Rohrbach, Lorenzo CavallaroMultimodal Model InterpretabilityUnified Multimodal Models

  2. EDA: Observe, Verify, and Assess for Evidence-Grounded Multimodal Depression Analysis

    Nov 17, 2026Changzeng Fu, Ming Shang, Xinyi Zhang +3Depression DetectionDepression Severity Estimation

  3. A Unified Interpretable Semantic Framework for Multi-Modal Fiber Sensing Under Data Scarcity

    Nov 11, 2026Khouloud AbdelliMultimodal Sensor FusionMultimodal Classification

  4. A Mechanistic Interpretability Study of an Astronomy Foundation Model

    Nov 11, 2026Rachel C. Zhang, Alberto Bietti, Wes Gurnee +2Multimodal Model InterpretabilityMechanistic Interpretability

  5. Structured Kernel Surrogate Models for Multimodal Attribution

    Nov 9, 2026Ziniu Zhang, Zhenshuo Zhang, Jianglin Lu +3Multimodal Model InterpretabilitySurrogate Modeling

  6. TIB: Sample-wise Tempered Information Bottleneck for Multimodal Attribution beyond Alignment Assumption

    Nov 8, 2026Qiming Huang, Peixi Liu, Jianbo JiaoMultimodal Model InterpretabilityInformation Bottleneck

  7. SWXAI: A Unified Multimodal Explainability Library for Space Weather Forecasting and Heliophysics

    Nov 7, 2026Benedict Antonio Mervyn, Anli Ji, Manolis K. Georgoulis +1Explainable Artificial IntelligenceMultimodal Model Interpretability

  8. Interpretable Context-Aware Multi-Label Retinal Classification with Dual-Hub Heterogeneous Graph Neural Network

    Nov 6, 2026Proshanto Chanda, Munib Mesinovic, Tingting Zhu +1Retinal Disease ClassificationMultimodal Graph Learning

  9. Glyph: An Interpretability Engine for AI-Driven Scientific Discovery from Clinical Data

    Nov 6, 2026Robbie Holland, Ashwin Kumar, Mateus A. Esmeraldo +19AI-Assisted Scientific ResearchMultimodal Model Interpretability

  10. Beyond the Readout: A Latent Numerical Frame Shared Across Modality and Operation

    Oct 27, 2026Oriol Juan Sabater, Alex Ferrando de las Morenas, Pau Rodríguez +1Multimodal Model InterpretabilityMultimodal Reasoning

  11. DocXEval: A Systematic Evaluation of Explainability in Multimodal Document Understanding

    Oct 27, 2026Saifullah Saifullah, David Dembinsky, Stefan Agne +2Multimodal Model InterpretabilityExplainability Evaluation

  12. Reliability-Aware Short-Term Glycemic Risk Prediction Using an Anchor-Based Comparative Framework of Sequence, Multimodal, and State-Updating Digital-Twin Models

    Oct 26, 2026Pir Bakhsh Khokhar, Carmine Gravino, Fabio PalombaDigital TwinsMultimodal Model Evaluation

  13. Faithful Explanations under Missing Modalities in Multimodal Dementia Prognosis

    Oct 26, 2026Firuz KamalovFaithfulness of Language Model ExplanationsMultimodal Model Interpretability

  14. All the Facts, Still the Wrong Answer: Integration Gaps in Vision-Language Models

    Oct 24, 2026Yuyang Dai, Yuxia Wang, Zhuohan XieVLM EvaluationCompositional Reasoning in VLMs

  15. A Broader Look into MLLM's Retrieval Heads

    Oct 23, 2026Jia Wang, Yuqing Li, Jiangnan Li +2Multimodal Large Language ModelsMultimodal IR

  16. CRANE: Post-Hoc Counterfactual Retrieval for Multimodal Classifiers

    Oct 22, 2026Franco Rugolon, Ioanna Miliou, Panagiotis PapapetrouCounterfactual ExplanationsMultimodal Model Interpretability

  17. I2^2-KAN: Interpretable Intra- and Inter-Modal Multi-Pathway Representation Learning for Multimodal Sentiment Analysis

    Oct 22, 2026Xiaoting Xie, Feng Qiao, Xianwei Zhang +5Multimodal Sentiment AnalysisMultimodal Fusion

  18. Encoder-Free Multimodal LLMs Learn Implicit Encoders

    Oct 16, 2026David Harwath, Desh Raj, Xavier Velez +5Multimodal Large Language ModelsMultimodal Model Interpretability

  19. Omni-Modal LLMs Know More Than They Say

    Oct 12, 2026Na Min An, Taekyung Kim, Jaemyung Yu +2Multimodal Large Language ModelsCross-Modal Alignment

  20. From Sparse Representations to Behavioral Insights for Multimodal Depression Assessment

    Oct 8, 2026Guimin Hu, Zihao Song, Jiachen Luo +2Depression DetectionMultimodal Model Interpretability

  21. Multimodal Safety Evaluation Should Measure Controllability Beyond Classification

    Oct 5, 2026Junhyeong Park, Hanwool Lee, DongGeon Lee +4Multimodal Model InterpretabilityMultimodal Model Evaluation

  22. Architecture-Dependent Fusion Pathways in MLLMs

    Oct 2, 2026Hebao Zhu, Dongxia WuMultimodal Large Language ModelsMultimodal Fusion

  23. Seeing What Should Be Heard: Diagnosing and Repairing Cross-Modal Shortcuts in Omni-Modal LLMs

    Sep 29, 2026Yueran Ma, Ronghao LinAudio QAMultimodal Large Language Models

  24. T-MoXAI: A Hierarchical Explainability Framework for Temporal Multimodal Data

    Sep 27, 2026Ali Inha, Mo Vali, Saaliha Vali +2Explainable Time Series ForecastingMultimodal Model Interpretability

  25. FCA-Guided Counterfactual Explanations for Multi-Modal Breast Cancer Diagnosis: A Framework Achieving Perfect Validity with Emergent Sparsity

    Sep 17, 2026Abdullahi Isa, Souley Boukari, Muhammad AliyuBreast CancerCounterfactual Explanations