Explaining multimodal LLMs via intra-modal token interactions

• URL: http://arxiv.org/abs/2509.22415v2
• Date: Wed, 01 Oct 2025 16:48:01 GMT
• Title: Explaining multimodal LLMs via intra-modal token interactions
• Authors: Jiawei Liang, Ruoyu Chen, Xianghao Jiao, Siyuan Liang, Shiming Liu, Qunli Zhang, Zheng Hu, Xiaochun Cao,
• Abstract summary: Multimodal Large Language Models (MLLMs) have achieved remarkable success across diverse vision-language tasks, yet their internal decision-making mechanisms remain insufficiently understood.<n>We propose enhancing interpretability by leveraging intra-modal interaction.
• Score: 55.27436637894534
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Multimodal Large Language Models (MLLMs) have achieved remarkable success across diverse vision-language tasks, yet their internal decision-making mechanisms remain insufficiently understood. Existing interpretability research has primarily focused on cross-modal attribution, identifying which image regions the model attends to during output generation. However, these approaches often overlook intra-modal dependencies. In the visual modality, attributing importance to isolated image patches ignores spatial context due to limited receptive fields, resulting in fragmented and noisy explanations. In the textual modality, reliance on preceding tokens introduces spurious activations. Failing to effectively mitigate these interference compromises attribution fidelity. To address these limitations, we propose enhancing interpretability by leveraging intra-modal interaction. For the visual branch, we introduce \textit{Multi-Scale Explanation Aggregation} (MSEA), which aggregates attributions over multi-scale inputs to dynamically adjust receptive fields, producing more holistic and spatially coherent visual explanations. For the textual branch, we propose \textit{Activation Ranking Correlation} (ARC), which measures the relevance of contextual tokens to the current token via alignment of their top-$k$ prediction rankings. ARC leverages this relevance to suppress spurious activations from irrelevant contexts while preserving semantically coherent ones. Extensive experiments across state-of-the-art MLLMs and benchmark datasets demonstrate that our approach consistently outperforms existing interpretability methods, yielding more faithful and fine-grained explanations of model behavior.

Related papers

• CrystaL: Spontaneous Emergence of Visual Latents in MLLMs [55.34169914483764]
  We propose CrystaL (Crystallized Latent Reasoning), a single-stage framework with two paths to process intact and corrupted images.<n>By explicitly aligning the attention patterns and prediction distributions across the two paths, CrystaL crystallizes latent representations into task-relevant visual semantics.<n>Experiments on perception-intensive benchmarks demonstrate that CrystaL consistently outperforms state-of-the-art baselines.
  arXiv  Detail & Related papers  (2026-02-24T15:01:30Z)
• Beyond Unimodal Shortcuts: MLLMs as Cross-Modal Reasoners for Grounded Named Entity Recognition [51.68340973140949]
  Multimodal Named Entity Recognition (GMNER) aims to extract text-based entities, assign them semantic categories, and ground them to corresponding visual regions.<n> MLLMs exhibit $textbfmodality bias$, including visual bias and textual bias, which stems from their tendency to take unimodal shortcuts.<n>We propose Modality-aware Consistency Reasoning ($bfMCR$), which enforces structured cross-modal reasoning.
  arXiv  Detail & Related papers  (2026-02-04T12:12:49Z)
• Video-STAR: Reinforcing Open-Vocabulary Action Recognition with Tools [41.993750134878766]
  Video-STAR is a framework that harmonizes contextual sub-motion decomposition with tool-augmented reinforcement learning for open-vocabulary action recognition.<n>Unlike prior methods that treat actions as monolithic entities, our approach innovatively decomposes actions into discriminative sub-motions for fine-grained matching.<n>Our method autonomously leverages external tools to prioritize sub-motion patterns without explicit supervision, transmitting from text-centric reasoning to visually grounded inference.
  arXiv  Detail & Related papers  (2025-10-09T17:20:44Z)
• UniAlignment: Semantic Alignment for Unified Image Generation, Understanding, Manipulation and Perception [54.53657134205492]
  UniAlignment is a unified multimodal generation framework within a single diffusion transformer.<n>It incorporates both intrinsic-modal semantic alignment and cross-modal semantic alignment, thereby enhancing the model's cross-modal consistency and instruction-following robustness.<n>We present SemGen-Bench, a new benchmark specifically designed to evaluate multimodal semantic consistency under complex textual instructions.
  arXiv  Detail & Related papers  (2025-09-28T09:11:30Z)
• Where MLLMs Attend and What They Rely On: Explaining Autoregressive Token Generation [59.40886078302025]
  Multimodal large language models (MLLMs) have demonstrated remarkable capabilities in aligning visual inputs with natural language outputs.<n>Yet, the extent to which generated tokens depend on visual modalities remains poorly understood.<n>We present a lightweight black-box framework for explaining autoregressive token generation in MLLMs.
  arXiv  Detail & Related papers  (2025-09-26T15:38:42Z)
• CLAMP: Contrastive Learning with Adaptive Multi-loss and Progressive Fusion for Multimodal Aspect-Based Sentiment Analysis [0.6961946145048322]
  This paper introduces an end to end Contrastive Learning framework with Adaptive Multi-loss and Progressive Attention Fusion.<n>The framework is composed of three novel modules: Progressive Attention Fusion network, Multi-task Contrastive Learning, and Adaptive Multi-loss Aggregation.<n> evaluation on standard public benchmarks demonstrates that CLAMP consistently outperforms the vast majority of existing state of the art methods.
  arXiv  Detail & Related papers  (2025-07-21T11:49:57Z)
• Attention Reallocation: Towards Zero-cost and Controllable Hallucination Mitigation of MLLMs [62.9348974370985]
  We propose attention reallocation (AttnReal) to mitigate hallucinations with nearly zero extra cost.<n>Our approach is motivated by the key observations that, MLLM's unreasonable attention distribution causes features to be dominated by historical output tokens.<n>Based on the observations, AttnReal recycles excessive attention from output tokens and reallocates it to visual tokens, which reduces MLLM's reliance on language priors.
  arXiv  Detail & Related papers  (2025-03-11T11:52:37Z)
• Multi-task Visual Grounding with Coarse-to-Fine Consistency Constraints [15.541287957548771]
  We propose a Coarse-to-fine Consistency Constraints Visual Grounding architecture.<n>It integrates implicit and explicit modeling approaches within a two-stage framework.<n>It significantly outperforms state-of-the-art REC and RIS methods by a substantial margin.
  arXiv  Detail & Related papers  (2025-01-12T04:30:13Z)
• Effective Context Modeling Framework for Emotion Recognition in Conversations [2.7175580940471913]
  Emotion Recognition in Conversations (ERC) facilitates a deeper understanding of the emotions conveyed by speakers in each utterance within a conversation.<n>Recent Graph Neural Networks (GNNs) have demonstrated their strengths in capturing data relationships.<n>We propose ConxGNN, a novel GNN-based framework designed to capture contextual information in conversations.
  arXiv  Detail & Related papers  (2024-12-21T02:22:06Z)
• What Do VLMs NOTICE? A Mechanistic Interpretability Pipeline for Gaussian-Noise-free Text-Image Corruption and Evaluation [16.033361754660316]
  Notice is the first Noise-free Text-Image Corruption and Evaluation pipeline for interpretability in Vision-Language Models (VLMs)<n>Our experiments on the SVO-Probes, MIT-States, and Facial Expression Recognition datasets reveal crucial insights into VLM decision-making.<n>This work paves the way for more transparent and interpretable multimodal systems.
  arXiv  Detail & Related papers  (2024-06-24T05:13:19Z)
• Cantor: Inspiring Multimodal Chain-of-Thought of MLLM [83.6663322930814]
  We argue that converging visual context acquisition and logical reasoning is pivotal for tackling visual reasoning tasks. We propose an innovative multimodal CoT framework, termed Cantor, characterized by a perception-decision architecture. Our experiments demonstrate the efficacy of the proposed framework, showing significant improvements in multimodal CoT performance.
  arXiv  Detail & Related papers  (2024-04-24T17:59:48Z)
• Sparsity-Guided Holistic Explanation for LLMs with Interpretable Inference-Time Intervention [53.896974148579346]
  Large Language Models (LLMs) have achieved unprecedented breakthroughs in various natural language processing domains. The enigmatic black-box'' nature of LLMs remains a significant challenge for interpretability, hampering transparent and accountable applications. We propose a novel methodology anchored in sparsity-guided techniques, aiming to provide a holistic interpretation of LLMs.
  arXiv  Detail & Related papers  (2023-12-22T19:55:58Z)
• Robust Latent Representations via Cross-Modal Translation and Alignment [36.67937514793215]
  Most multi-modal machine learning methods require that all the modalities used for training are also available for testing. To address this limitation, we aim to improve the testing performance of uni-modal systems using multiple modalities during training only. The proposed multi-modal training framework uses cross-modal translation and correlation-based latent space alignment.
  arXiv  Detail & Related papers  (2020-11-03T11:18:04Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.