Multimodal Model Interpretability

Multimodal large language models (MLLMs) generate responses autoregressively, integrating visual and linguistic information in an evolving context. Prior work on interpretability has focused on individual layers and circuits (where), leaving the token-level dynamics of multimodal computation during generation (when) underexplored. We address this gap and study attention shifts as per semantic role; tracking model attention to image, text, instruction, and previously generated tokens, One Token at a Time (OTaT). We introduce multimodal tasks that require explicit switching between visual and textual context within a single response. Across two mainstream model families and four open-weight MLLMs of varying sizes, we establish consistent patterns: attention to image peaks at tokens requiring image-derived information, instruction tokens are revisited during task transitions, and attention to previously generated tokens increases as the generation progresses. Causal attention blocking interventions validate the functional role of these trends. We profile model behavior under disrupted attention and observe responses falling back to language priors, or exhibiting cross-modal leakage, denial, or recovery. Finally, informed of the attention dynamics through our novel analysis, we propose a simple test-time intervention to boost attention to the relevant modality at the right time, significantly improving multimodal task performance.

Despite the high accuracy of EEG-based emotion recognition, existing models remain opaque "black boxes", lacking semantic grounding between abstract neural features and human-interpretable states. In this paper, we reframe EEG explainability as a cross-modal generation task, shifting the paradigm from feature attribution to behavioral visualization. We introduce Facial Emoji Proxy Modeling, a novel framework that translates high-dimensional EEG signals into identity-anonymized facial emojis. Guided by the neuroscientific inspiration of neural-facial association, this approach grounds neural representations in the manifold of observable facial dynamics. Technically, our framework integrates FMENet, a specialized backbone modeling expression-relevant spatial synergies, and the Facial Emoji Learning Branch (FELB), which treats emoji reconstruction as a structured semantic regularizer. Extensive experiments on EAV and MMER benchmarks demonstrate that our method achieves state-of-the-art accuracy among EEG-only models. Crucially, it generates semantically faithful facial animations that provide a transparent, privacy-preserving window into the brain's emotional evolution, effectively allowing users to "see the emotion" directly from neural signals. Code is available at https://github.com/xian-sh/SeeEmotion

The interaction between brain structure and genetic influences is key to understanding neuropsychiatric disorders. However, most large-scale datasets are unimodal, providing either neuroimaging or genetics data. We propose CALM, a framework that learns interpretable associations between brain ROIs and genetic pathways from completely disjoint populations. CALM aligns the two modalities in a shared latent space via linear projections that simultaneously match the class-conditional latent distributions and ensure group separability. These projections provide interpretable pathway--ROI associations. When trained on unimodal imaging and genetics datasets, CALM generalizes to an unseen paired dataset, outperforming several state-of-the-art methods and ablation baselines. We also demonstrate stability of the learned associations against a paired baseline. Our experiments on autism spectrum disorder reveal immune and metabolic pathways linked to specific cortical regions and are consistent with established literature. Thus, CALM opens the door to leveraging large unimodal repositories for studying cross-modal interactions in brain disorders across disparate datasets.

As grounded QA systems are increasingly deployed in AI assistants, accurately attributing generated answers to evidence is critical for user trust and model safety. While unimodal attributions have been explored in depth, the multimodal setting remains relatively under-researched. As a result, we introduce MultAttnAttrib, a training-free attribution-generation method that leverages a model's prefill pass, selected attention heads, and calibrated thresholds to locate source evidence within a document. To establish baseline results for the method, we introduce MultAttrEval, a complementary benchmark dataset annotated with fine-grained, ground-truth attributions for answer components grounded in multimodal source documents. To our knowledge, this is the first evaluation dataset designed specifically for multimodal attribution in long-form documents. Experimental results show that MultAttnAttrib consistently outperforms a variety of attribution-generation methods, including several strong prompting-based approaches and matches the latest frontier models such as GPT 5.4. Our method not only substantially improves attribution accuracy for both unimodal and multimodal attribution types, but also produces attributions at up to one-seventh of the direct inference latency compared to prompting on the same base model.

Multimodal graph foundation models aim to learn reusable knowledge from graphs enriched with text, images, attributes, and relational topology, thereby supporting diverse graph-centric and modality-centric tasks. In practice, however, such multimodal graphs are often distributed across decentralized clients, where raw contents and local structures cannot be centrally shared due to privacy constraints. This motivates federated multimodal graph foundation learning, which requires not only transferable representation learning but also intrinsic semantic traceability under strict data isolation. Existing methods usually exchange or store knowledge through parameters, prototypes, embeddings, or compact codebooks, which support optimization and transfer but do not explicitly expose how modality evidence, node semantics, and topology context jointly support predictions. To bridge this gap, we propose FedLAB, a traceable semantic codebook framework that organizes multimodal graph knowledge into typed hierarchical codebooks for modality evidence, node semantics, and topology context. FedLAB further refines these trace units through federated semantic barycenter pre-training while keeping raw multimodal contents and graph structures local. Extensive experiments on 10 benchmarks and 6 downstream tasks show that FedLAB improves over state-of-the-art baselines by up to 7.53%, while preserving a native semantic trace interface.

This paper offers an in-depth analysis of non-sequential multimodal sentence-level embeddings, with a particular focus on the SONAR model. We demonstrate that certain embedding dimensions are sensitive to perturbations and can serve as indicators of decoding anomalies. By leveraging the consistency between successive encoding and decoding, we successfully build an accurate detector. Additionally, we explore modifying specific dimensions of interest to attempt to correct them. This work underscores the importance of understanding and analyzing the embeddings themselves to enhance the reliability of multimodal representations.

Encrypted traffic classification has achieved strong performance, but its decision process remains difficult to interpret. Existing methods usually combine flow statistics, packet sequences, and byte-level representations into opaque latent features, making it unclear which type of evidence actually drives the prediction. In this paper, we propose Traffic-CBM, a structurally interpretable multimodal framework for encrypted traffic classification. Instead of directly fusing heterogeneous traffic signals into a black-box representation, Traffic-CBM organizes them into a unified hierarchical concept space. These concepts are not manually annotated semantic attributes; rather, they are scalar evidence summaries constrained by predefined traffic evidence groups. More specifically, grouped flow statistics are mapped to statistical concepts, dedicated temporal encoders learn temporal concepts from disjoint feature subspaces, and byte-level evidence is further organized into packet-level and cross-packet concepts. This design turns heterogeneous traffic evidence into an explicit concept representation and makes different levels of traffic evidence easier to analyze. We evaluate Traffic-CBM on multiple encrypted traffic benchmarks. Results show that it achieves competitive and balanced classification performance while providing a clearer structural interpretation interface than conventional end-to-end fusion models. Further analyses suggest that the learned concept space is actively used in the prediction process and provides a clearer structural explanation of multimodal traffic evidence.

Mixture-of-Experts (MoE) architectures have recently been extended with role-based mechanisms for interpretability. This is typically done by assigning semantic roles to individual expert components, for example roles like synergy, redundancy, and uniqueness in multimodal settings. However, whether such structural role decomposition preserves explanation faithfulness of the overall architecture remains largely underexplored. We hypothesize that inter-expert representation overlap weakens effective role separation and degrades attribution-based faithfulness, even when semantic roles are explicitly defined. To address this limitation, we introduce representation-level decorrelation regularization to explicitly reduce inter-expert similarity in latent space. Using representation decorrelation objectives, we encourage clearer specialization among experts by minimizing representation overlap. Our experiments show that across multiple multimodal benchmarks, this separation consistently improves explanation faithfulness, as measured by comprehensiveness, sufficiency, and their Area Over the Perturbation Curve (AOPC) summaries, while preserving task performance. We further show that these improvements are not limited to role-based architectures such as Interpretable Multimodal Interaction-aware MoE (I2MoE). Similar trends are observed in a standard sparse MoE baseline, suggesting that representation-level separation may provide a more general mechanism for enhancing explanation faithfulness in MoE systems. Overall, our findings suggest that structural role decomposition alone may be insufficient to guarantee faithful explanations and that representation-level separation helps improve explanation faithfulness. To support reproducibility, the source code and supplementary material are publicly available at https://github.com/dut0817/FL-I2MoE_Decor.

Tailoring stroke rehabilitation requires assessing how movements are organized, not merely if they succeed. Currently, this assessment is a rate-limiting bottleneck. Instruments like the Action Research Arm Test (ARAT) compress rich behavioral observations into single ordinal endpoints, discarding the movement-quality details that distinguish recovery from compensation. Automated alternatives typically chase accuracy on noisy, single-observer labels to output opaque scores - a technology-centric approach that rarely reaches clinical practice. To address this, we present xAARA: an engine designed to augment rather than replace clinical judgment. From multi-view video, xAARA returns ARAT assessments with calibrated uncertainty and explanations across task, movement-phase, and movement-quality levels. Treating clinical scoring as an ill-posed inference problem, xAARA composes 692 calibrated multimodal models via a Dynamic Bayesian Network with entropy-based gating. It qualifies results against clinical validity rules and defers low-confidence cases. In 105 stroke survivors (788 exercises), xAARA achieved 94.2% task accuracy (Cohen's kappa=0.934) and 81.3% movement-phase accuracy (kappa=0.727), reducing predictive uncertainty by 96.1% compared to single-clinician scoring. For subjective cases, it matched at least one rater 100% of the time and never returned out-of-range scores. Four independent clinicians validated the assessments and indicated willingness to adopt the system. We argue that principled uncertainty quantification and clinician-aligned explainability are the critical bridges moving automated assessment from technical demonstration to a deployable clinical tool.

Foundation models have achieved remarkable performance across medical question answering, imaging, and electronic health record (EHR) tasks, yet reliable clinical deployment remains challenging due to limited interpretability, vulnerability to distribution shift, and weak alignment with clinician reasoning. We argue that these limitations arise because existing approaches prioritize representation learning while treating patient comparison as an emergent property rather than a primary source of clinical evidence. To address this gap, we propose CAFM, a Cohort-Anchored Foundation Model framework that elevates patient cohorts to a first-class object throughout the learning pipeline. The framework consists of four stages: deviation-aware data curation, cohort-conditioned pretraining, multimodal cohort alignment, and clinician-in-the-loop refinement. Together, these stages improve data quality, organize representations around clinically meaningful cohort structure, preserve modality-specific relationships, and support auditable clinical decision-making. The framework is compositional and can augment existing EHR foundation models without modifying their underlying encoders. We illustrate CAFM through four clinical case studies spanning acute kidney injury prediction, cardiovascular risk stratification from electrocardiograms, optic neuropathy triage from orbital imaging, and electroretinogram-grounded report generation. We further present five empirically testable hypotheses and identify open challenges in data quality, irregular temporality, multimodal learning, distribution shift, and evaluation beyond predictive accuracy. We argue that explicitly anchoring foundation models to patient cohorts provides a principled path toward trustworthy clinical AI.

Concept Bottleneck Models (CBMs) enhance the interpretability of deep learning networks by aligning the features extracted from images with natural concepts. However, existing CBMs are constrained in their ability to generalize beyond a fixed set of predefined classes and the risk of non-concept information leakage, where predictive signals outside the intended concepts are inadvertently exploited. In this paper, we propose Multimodal Concept Bottleneck Model (MM-CBM) to address these issues and extend CBMs into CLIP. MM-CBM utilizes dual Concept Bottleneck Layers (CBLs) to align both the image and text embeddings into interpretable features. This allows us to perform new vision tasks like zero-shot classification or image retrieval in an interpretable way. Compared to existing methods, MM-CBM achieves up to 51.26% accuracy improvement on average across four standard benchmarks. Our method maintains high accuracy, staying within ~5% of black-box performance while offering greater interpretability.

This article presents a complementary approach for integrating multimodal medical data in cancer classification, based on state space models represented by the Mamba architecture. To this end, a mixed multimodal fusion architecture, called Mixed Fusion, was employed and developed to enhance the interpretability of the decision-making process. The proposed approach explores two variants of Mamba: one dedicated to visual processing, responsible for classifying the lesion image and generating probabilities associated with the target classes, and another focused on tabular processing, which uses these probabilities together with clinical and/or sociodemographic data to produce the final diagnosis. The experiments were conducted on two medical datasets: PAD-UFES-20, composed of clinical images and information associated with skin lesions, and NDB-UFES, consisting of histopathological images and sociodemographic data related to oral cancer. The results indicate slightly lower performance in balanced accuracy, compared with Transformer-based approaches, on PAD-UFES-20, and superior performance on NDB-UFES. Additionally, substantial gains were observed in the recall metric. Furthermore, the adoption of the Mixed Fusion architecture enables the application of the Shapley Additive Explanations (SHAP) method, increasing the interpretability of the results. These findings indicate that Mamba-based models constitute a suitable alternative for multimodal classification in medical data, especially in scenarios in which sensitivity is a relevant requirement.

Multimodal deception detection is critical for identifying fraudulent intentions, yet existing approaches predominantly rely on end to end black--box paradigms. These methods suffer from a severe lack of interpretability failing to provide transparent reasoning trajectories and struggling to explicitly capture the subtle, cross modal inconsistencies inherent in deceptive behaviors. To transcend these limitations, we propose ThinkDeception, a novel and interpretable multimodal deception detection framework. As a pioneering effort, it introduces Multimodal Large Language Models (MLLMs) into this domain, transforming deception detection from a traditional binary classification task into an explicit cognitive reasoning process. Facilitated by the first meticulously annotated step--by--step multimodal Chain of Thought (CoT) dataset, we develop a foundational model, ThinkDeception Base, empirically validating the critical role of modal inconsistency in decoding deception. Building upon this foundation, our core innovation lies in proposing Visual-Audio Consistency Group Relative Policy Optimization(VAC--GRPO) equipped with a progressive training strategy. Distinct from standard GRPO, we stratify the training data into four progressive difficulty tiers, guiding the model through a psychologically grounded easy--to--hard cognitive transition. By innovatively coupling this dynamic curriculum scheduler with a multi dimensional, process aware reward mechanism and a reflective learning paradigm, we significantly elevate the model's overall reasoning quality. Extensive experiments on mainstream benchmarks demonstrate that ThinkDeception establishes a new SOTA, significantly outperforming existing methods in both detection accuracy and rationale quality. Ultimately, this work successfully drives the field of deception detection toward interpretable, multimodal cognitive reasoning.

Multimodal Large Language Models (MLLMs) have demonstrated strong performance on vision-language tasks, yet their internal visual representations remain difficult to interpret. Sparse Autoencoders (SAEs) provide a scalable way to decompose dense model activations into sparse, interpretable features. However, existing SAE architectures primarily recover flat feature dictionaries and are less suited for explicit multi-level concept organization. In this paper, we introduce a cascaded sparse autoencoder architecture, dubbed SAE++, for learning hierarchical visual concepts in MLLMs. Rather than nesting or stacking SAE sparse activation codes, SAE++ trains a second-level SAE directly on the decoder weights of the first-level SAE, treating learned low-level feature directions as inputs for higher-level abstraction. This design enables SAE++ to learn "concepts of concepts" while avoiding drawbacks from the shared-prefix coupling of nesting, Matryoshka-style hierarchies and the bottlenecks of naively stacked SAEs. Experiments across Qwen3-VL, Gemma-3, and LLaVA on multiple visual datasets show that SAE++ improves interpretability in terms of hierarchical concept coherence over state-of-the-art SAE baselines. Results on concept steering further demonstrate that the learned concept groups support effective group-level interventions in MLLM outputs. Code is available at https://github.com/Wang-ML-Lab/sae-plus-plus.

Mechanistic interpretability seeks to explain neural network behavior by decomposing model computations into interpretable features and circuits. While transcoder-based circuit tracing has recently enabled detailed causal analyses of large language models, multimodal diffusion transformers for image generation remain comparatively opaque. We still lack tools for understanding how semantic information propagates across denoising steps and how text and image representations interact within double-stream MM-DiT architectures. Existing methods provide only partial insight: attention maps expose a limited view of token interactions, while sparse autoencoders can discover interpretable features but do not directly reveal how these features are transformed and composed through nonlinear MLP layers. In this work, we extend transcoder-based circuit tracing to multimodal diffusion transformers. We train timestep-conditioned transcoders that faithfully approximate the input-output behavior of MLP sublayers in FLUX.1[schnell]. By replacing MLPs with transcoders and linearizing the remaining computation, we obtain exact feature-to-feature attribution and recover compact, interpretable circuits. Empirically, our transcoders match or slightly outperform sparse autoencoders on the sparsity-faithfulness tradeoff. The resulting circuits reveal mechanisms underlying attribute binding and cross-stream semantic propagation, and provide causal explanations for systematic generation errors. Moreover, circuit-guided interventions are substantially more precise and effective than standard SAE-based steering. Our results demonstrate that transcoder-based circuit analysis is feasible for state-of-the-art diffusion transformers and provides a powerful framework for understanding and controlling multimodal generative models. The code is available at https://github.com/Artalmaz31/DifFRACT

Deception detection is a critical and highly challenging task within affective computing and behavioral analysis. Existing deep learning methods typically treat this task as a straightforward classification problem; however, this black-box approach lacks interpretability and fails to capture the complex logical deduction processes utilized by human experts when identifying lies. While Multimodal Large Language Models (MLLMs) have shown potential, applying them effectively requires a bridge between low-level audiovisual cues and high-level logical reasoning. In this paper, we propose DeceptionX, a novel MLLM framework that shifts the paradigm of deception detection from black-box classification to an interpretable Observe-Think-Summarize reasoning process. To address the scarcity of high-quality reasoning data, we first constructed DeceptChain, a high-quality dataset developed through a human-in-the-loop process. This dataset synthesizes fine-grained visual and auditory evidence (such as micro-expressions and vocal tremors) into structured chain-of-thought reasoning data. Furthermore, we propose a three-stage training pipeline and a Discrepancy-Aware Redundancy Elimination~(DARE) strategy for DeceptionX to further enhance the model's generalization capabilities. Extensive experiments demonstrate that DeceptionX not only outperforms existing MLLM baselines and state-of-the-art methods on standard real-world benchmarks but also provides transparent, expert-level reasoning paths, bridging the critical gap between accuracy and interpretability in multimodal deception detection.

Multimodal Large Language Models (MLLMs) can listen and see, but how do audio and visual signals actually travel through the network to shape an answer? Despite their growing role in research and real-world applications, the internal pathways through which audio and visual tokens influence the final prediction remain poorly understood. In this study, we examine audio-visual information flow inside Audio-Visual Large Language Models (AVLLMs), tracing how AVLLMs route, utilize, and integrate audio and visual information across two input configurations, audio-visual video and multiple interleaved audio-visual items. We find that for audio-visual video, AVLLMs follow the sequential information flow pathway established for VLMs and VideoLLMs, with audio and visual contribution flowing along this pathway in proportion to the task's reliance on each modality. In settings with multiple interleaved audio-visual items, this routing shifts to different parallel streams. Furthermore, we demonstrate that audio-visual and other token types can be discarded once their information is transferred to LLM, with minimal impact on the model's prediction or even slight improvement, generalizing across multiple tasks and datasets, enabling more efficient inference. These findings hold across multiple models and scales, Qwen2.5-Omni and Video-SALMONN2 Plus at 3B and 7B scales, leading to hypotheses on why these flow structures emerge. Together, these results deliver the first coherent picture of how AVLLMs orchestrate sound and sight inside the network and lay the groundwork for the next wave of interpretability, design, and efficiency advances in audio-visual and broader MLLMs.

As Vision-Language Models are increasingly deployed in safety-critical applications, the trustworthiness of their explanations becomes crucial. Explainable AI (XAI) methods for Vision-Language Models often suffer from semantic hallucination, where attribution maps highlight prominent image regions even when prompted with incorrect text descriptions (e.g., highlighting a dog when prompted ``cat''). Although this problem is widespread, a formal mathematical analysis of XAI methods and CLIP embeddings is largely missing in the literature. We demonstrate that this phenomenon is not specific to a single architecture but is a fundamental consequence of Linear Semantic Leakage in high-dimensional embedding spaces. We propose a unified theoretical framework, Linear Semantic Attribution (LSA), which generalizes across discriminative methods. We introduce OSP, a geometric intervention that utilizes the residual property of OMP to disentangle unique semantic signals from shared concepts. We prove theoretically and demonstrate empirically that OSP minimizes hallucination by orthogonalizing the query vector against distractor concepts, rendering the attribution model blind to shared features while preserving fidelity for correct prompts. Our code is available at: https://github.com/emirhanbilgic/Orthogonal-Semantic-Projection

Multimodal language models are typically evaluated through external behavior: selecting the correct image--text match, rejecting unsupported captions, or answering visual queries correctly. However, correct behavior alone does not show that the model's internal decision state remains stable under controlled semantic stress. We study this gap through S$^3$E (Structured Semantic Stress Evaluation), a framework for analyzing behavior-internal decoupling in multimodal language models. S$^3$E uses a positive-anchored A/B forced-choice setup in which an image-supported caption is contrasted against semantic stress candidates under both original and swapped option orders, while hidden states are extracted at the pre-answer decision state. We focus on strict-correct trials, where the model consistently selects the correct caption across both orders. Rather than treating arbitrary hidden-state variation as evidence of instability, we measure whether semantic-conflict candidates induce excess decision-state displacement relative to meaning-preserving controls. Across Qwen3VL, Gemma3, and InternVL3, semantic stress consistently produces positive selected-layer excess displacement over lexical controls despite correct forced-choice behavior, while comparisons against random negatives are model-dependent. We interpret this as a scoped decision-state stress-sensitivity signal rather than evidence of downstream failure or hallucination. Our results suggest that forced-choice correctness alone is not a sufficient certificate of invariant internal decision geometry.

With the increasing development of Vision-Language Models, it becomes imperative that their predictions are readily explainable to relevant stakeholders. However, the field of explainability has not kept pace with the multimodal surge. While recent Multimodal Explainable AI (MxAI) methods generate explanations to attribute the interaction between different modalities, current evaluation protocols lack the ground truth required to distinguish between true cross-modal reasoning (e.g., spatial composition) and shallow cross-modal shortcuts (e.g., Bag-of-Words attribute matching). It remains unknown whether MxAI methods faithfully capture synergistic interactions or merely hallucinate reasoning on models acting as simple feature detectors. In this paper, we introduce GridVQA-X, the first diagnostic framework specifically designed to evaluate cross-modal explainability. Unlike natural datasets, GridVQA-X leverages a closed-world synthesis logic to generate unique, mathematically guaranteed explanations. We utilize this controlled environment to train paired ground-truth models on identical architectures: $M_{\text{pure}}$, which learns robust spatial-relational reasoning and $M_{\text{spur}}$, which is structurally forced to rely on cross-modal shortcuts. This behavioral divergence creates a rigorous testbed: a faithful explainer must report distinct reasoning pathways for each model. Our findings reveal that widely used methods fail to distinguish between models relying on genuine spatial-relational reasoning and those exploiting cross-modal shortcuts, highlighting a critical gap in capturing true cross-modal synergy and misrepresenting how multimodal models actually make decisions.

Understanding modality interaction in multimodal large language models (MLLMs) is central to reliable deployment. We introduce Partial Information Decomposition (PID) as a decision-level framework that separates unique, redundant, and synergistic contributions of sensory and linguistic inputs, beyond representation alignment and outcome-based evaluation. Across vision--language benchmarks, PID reveals recurring modality-use profiles: reasoning and grounding-oriented tasks tend to exhibit high synergy, whereas expert and knowledge-oriented tasks show stronger language-unique reliance. These profiles generalize across model families and predict sensitivity to modality-level interventions. We further extend PID to tri-modal systems with Sensory PID, treating language as a control variable to decompose video--audio information gain. Applied to omni-modal models, Sensory PID reveals a sensory synergy bottleneck dominated by visual information even on audio--visual fusion tasks. Finally, PID-guided reweighting provides initial evidence for improving multimodal reasoning and grounding performance.

This work presents MLLM-Microscope, a novel system designed for analyzing the hidden representations within Multimodal Large Language Models (MLLMs). Our system evaluates the linearity, intrinsic dimension, and anisotropy of multimodal token embeddings across transformer layers. Utilizing the ScienceQA dataset, we evaluate two state-of-the-art MLLMs, LLaVA-NeXT and OmniFusion. We find that both the main and residual streams for tokens of both modalities exhibit highly linear behaviors across transformer layers. However, LLaVA-NeXT's image tokens reveal a slight decline in linearity, whereas OmniFusion's remain consistent. Image token dimensions in OmniFusion remain consistently higher across layers compared to LLaVA-NeXT. Also, the OmniFusion's anisotropy is observed to stay consistently low throughout the layers. These findings suggest that the inner workings of MLLMs highly depend on the nature of modality fusion performed before passing the token sequence into LLM. This and other new potential insights obtainable from our system are surely capable of enhancing our understanding of the inner workings of MLLMs, informing future model design and optimization.

Healthcare models are transitioning from unimodal prediction toward multimodal reasoning over heterogeneous diagnostic inputs. In computational pathology, for complex tumor subtypes where morphology alone can be challenging to distinguish, pathology reports and molecular measurements may provide additional diagnostic evidence alongside whole-slide images, yet existing models often fail to clarify how diverse signals assemble into recognizable diagnostic concepts. We propose ConceptM$^3$oE (Concept Multimodal MoE), which embeds concept formation directly within interaction-aware mixture-of-experts (MoE) pathways. The architecture decomposes evidence into modality-specific, redundant, and synergistic experts, which are then projected into structured concept bottlenecks mapping latent features to a hierarchy of morphology and biomarker concepts. To prevent the information loss typical of interpretable bottlenecks, we utilize residual pathways within each expert to allow task-relevant signals to flow both through the concepts and directly to the final task prediction, so that high performance is maintained alongside interpretability. Across an institutional pediatric brain tumor cohort and a public glioma cohort, the framework delivers competitive performance to unconstrained models while producing reasoning traces validated by an independent neuropathologist. In data-limited regimes, ConceptM$^3$oE improves limited-data performance, increasing macro-F1 from 56.41% to 66.70% at small training sizes compared to non-concept-informed baselines, while also showing faster training convergence consistent with the regularizing effect of concept learning. This work offers a scalable path toward high-performance medical AI that is inherently verifiable and better aligned with the complex decision-making of clinical practice.

Generative Vision-Language Models (VLMs) perform well on multimodal reasoning, but how visual inputs are transformed to text remains poorly understood. Existing interpretability work on VLMs uses Sparse Autoencoders (SAEs), which decompose static residual representations and miss the functional updates that drive cross-modal interaction. We adopt a function-centric framework based on Transcoders, sparse approximations of MLP sublayers that act as a causal proxy for layer-wise computation. Applied to Gemma 3-4B-IT, the framework decomposes the model into interpretable computational pathways linking image patches to directions in token generation. Transcoder attributions produce stronger and more stable effects on visually grounded tokens under patch ablation than SAE attributions, and align better with semantically relevant image regions. A False Visual Grounding counterfactual analysis confirms that the recovered pathways are specific to vision-language interaction.Finally, we perform a structural analysis of hallucinated generations, by extracting graph-based indicators from circuit traces produced by the transcoders. A logistic classifier over these mechanistic graph features predicts hallucinations at AUC $0.68$. These results show that function-centric circuit decomposition yields interpretable and predictive accounts of multimodal computation in VLMs.

Vision-Language Models (VLMs) map complex visual inputs to semantic spaces, but interpreting the cross-modal reasoning of VLMs currently relies on post-hoc explainers evaluated via unimodal perturbation metrics. We expose a limitation in this paradigm: because multimodal datasets contain language priors and modality biases, VLMs frequently exhibit cross-modal redundancy, allowing them to answer visual queries using text alone. Consequently, unimodal metrics penalize faithful explainers, triggering an evaluation collapse where visual and textual rankings fundamentally contradict each other. %(Kendall's $τ= -0.06$). To resolve this, we introduce Synergistic Faithfulness ($\mathcal{F}_{syn}$), a scalable metric rooted in the Shapley Interaction Index that strictly isolates the joint Harsanyi dividend between modalities, serving as a highly accurate surrogate ($ρ= 0.92$) while achieving a $24\times$ computational speedup. Evaluating 8 distinct XAI methods across 3 VLM architectures and 3 benchmark datasets, reveals that explainers proposed for VLMs heavily over-index on visual salience and significantly underperform adapted attention-based methods in capturing true cross-modal synergy. By decoupling visual plausibility from cross-modal faithfulness, this work provides a rigorous evaluation framework required to safely audit VLM reasoning in high-stakes deployments.

While interpretable prototype networks offer compelling case-based reasoning for clinical diagnostics, their raw continuous outputs lack the semantic structure required for medical documentation. Bridging this gap via standard Retrieval-Augmented Generation (RAG) routinely triggers ``retrieval sycophancy,'' where Large Language Models (LLMs) hallucinate post-hoc rationalizations to align with visual predictions. We introduce ProtoMedAgent, a framework that formalizes multimodal clinical reporting as an iterative, zero-gradient test-time optimization problem over a strict neuro-symbolic bottleneck. Operating on a frozen prototype backbone, we distill latent visual and tabular features into a discrete semantic memory. Online generation is strictly constrained by exact set-theoretic differentials and a reflective Scribe-Critic loop, mathematically precluding unsupported narrative claims. To safely bound data disclosure, we introduce a semantic privacy gate governed by $k$-anonymity and $\ell$-diversity. Evaluated on a 4,160-patient clinical cohort, ProtoMedAgent achieves 91.2% Comparison Set Faithfulness where it fundamentally outperforms standard RAG (46.2%). ProtoMedAgent additionally leverages a binding $\ell$-diversity phase transition to systematically reduce artifact-level membership inference risks by an absolute 9.8%.

We present Supervised Deep Multimodal Matrix Factorization (SD3MF), an interpretable framework for integrative brain network analysis that generalizes Symmetric Nonnegative Matrix Tri-Factorization (SNMTF) from unsupervised single-graph clustering to supervised prediction over populations of multimodal graphs. SD3MF learns deep hierarchical factorizations for each modality together with a shared latent representation that aligns subjects across views. An encoder-decoder formulation jointly optimizes graph reconstruction and supervised prediction, while adaptive weights enable data-driven multimodal fusion. By representing each subject through community-level interaction matrices, the model yields interpretable and discriminative features. Experiments on multimodal connectome datasets show that SD3MF consistently outperforms strong deep learning baselines such as CNNs and GNNs, while enabling biologically interpretable insights. Code for reproducibility is available at: https://github.com/amjadseyedi/SD3MF.

Audio-visual large language models (AVLLMs) have recently emerged as a powerful architecture capable of jointly reasoning over audio, visual, and textual modalities. In AVLLMs, the bidirectional interaction between audio and video modalities introduces intricate processing dynamics, necessitating a deeper understanding of their internal mechanisms. However, unlike extensively studied text-only or large vision language models, the internal workings of AVLLMs remain largely unexplored. In this paper, we focus on cross-modal information flow between audio and visual modalities in AVLLMs, investigating where information derived from one modality is encoded within the token representations of the other modality. Through an analysis of multiple recent AVLLMs, we uncover two common findings. First, AVLLMs primarily encode integrated audio-visual information in sink tokens. Second, sink tokens do not uniformly hold cross-modal information. Instead, a distinct subset of sink tokens, which we term cross-modal sink tokens, specializes in storing such information. Based on these findings, we further propose a simple training-free hallucination mitigation method by encouraging reliance on integrated cross-modal information within cross-modal sink tokens. Our code is available at https://github.com/kaistmm/crossmodal-hub.

How can an arbitrary attribution method be generalized from first principles to capture interactions? We answer this with the metagame, a conceptual framework for quantifying second-order interaction effects of model explanations. We cast the attribution value $φ_i$ of feature $i$ as a cooperative game among the other features and compute its Shapley value, which measures how much feature $j$ influences the attribution of $i$, yielding the directional meta-attribution $\varphi_{j \to i}$. By decomposing attribution itself rather than the model directly, meta-attributions extend any gradient- or attention-based method to interactions, uniting removal-based perturbations with model internals. Theoretically, we prove that meta-attributions sum to the first-order attribution they explain, a hierarchical decomposition that Shapley interactions and integrated Hessians turn out to perform implicitly. Empirically, we demonstrate that meta-attributions deliver insights across diverse interpretability applications: (i) quantifying token interactions in instruction-tuned language models, (ii) explaining cross-modal similarity in vision-language encoders, and (iii) interpreting text-to-image concepts in multimodal diffusion transformers.

Achieving resilient and sustainable cities requires scalable approaches to decarbonising residential buildings, which account for about 20% of UK greenhouse gas emissions and 25% of energy-related emissions in the European Union. Energy Performance Certificates (EPCs) support regulation and retrofit planning, but their reliance on on-site inspections limits timely city-scale assessment. This study introduces a gated multimodal model to predict Standard Assessment Procedure (SAP) energy efficiency and Environmental Impact (EI) scores by integrating EPC tabular variables, assessor-written free text, and Geographic Information System (GIS)-derived spatial features describing footprint geometry, height, area, and orientation. Sample-wise gating learns property-specific modality weights, while an auxiliary band classification head stabilises training. In a Westminster, London case study, the model predicts SAP and EI scores with MAEs of 4.03 and 4.76 points and R2 values of 0.757 and 0.748, respectively, achieving a mean MAE of 4.39. Ablation results show that full multimodal fusion outperforms unimodal and bimodal baselines for both score prediction and band-level classification. Interpretability analyses provide decision-relevant evidence: gating weights indicate strong reliance on assessor text; SHAP highlights main fuel, built form, and construction age band; text occlusion prioritises roof and wall fields; and spatial attribution is dominated by height and footprint area, with sensitivity to footprint shape. The validated framework is further applied to retrofit scenarios for wall insulation, roof insulation, and window glazing upgrades, indicating projected improvements in SAP, EI, annual energy cost, and equivalent CO2 emissions. Overall, the framework provides scalable property-level evidence for retrofit screening, intervention prioritisation, and net-zero housing transitions.