DEX-AR: A Dynamic Explainability Method for Autoregressive Vision-Language Models

2026-03-06

AI Summary

Existing interpretability methods struggle to capture the complex multimodal interactions in autoregressive vision-language models during token-by-token generation. To address this, this work proposes a dynamic interpretability framework that introduces a dynamic attention head filtering mechanism to identify vision-sensitive attention heads and employs a sequence-level strategy to distinguish visually grounded tokens from purely linguistic ones. Building on the token-wise generation process, the method computes hierarchical gradients to produce both token-level and sequence-level 2D saliency maps that localize critical image regions. Experiments on ImageNet, VQAv2, and Pascal VOC demonstrate that the proposed approach significantly outperforms existing techniques in perturbation tests and segmentation-based evaluation metrics.

Technology Category

• Computer Vision: Multi-modal Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

As Vision-Language Models (VLMs) become increasingly sophisticated and widely used, it becomes more and more crucial to understand their decision-making process. Traditional explainability methods, designed for classification tasks, struggle with modern autoregressive VLMs due to their complex token-by-token generation process and intricate interactions between visual and textual modalities. We present DEX-AR (Dynamic Explainability for AutoRegressive models), a novel explainability method designed to address these challenges by generating both per-token and sequence-level 2D heatmaps highlighting image regions crucial for the model's textual responses. The proposed method offers to interpret autoregressive VLMs-including varying importance of layers and generated tokens-by computing layer-wise gradients with respect to attention maps during the token-by-token generation process. DEX-AR introduces two key innovations: a dynamic head filtering mechanism that identifies attention heads focused on visual information, and a sequence-level filtering approach that aggregates per-token explanations while distinguishing between visually-grounded and purely linguistic tokens. Our evaluation on ImageNet, VQAv2, and PascalVOC, shows a consistent improvement in both perturbation-based metrics, using a novel normalized perplexity measure, as well as segmentation-based metrics.

Problem

• Vision-Language Models
• Explainability
• Autoregressive Models
• Token-by-token Generation
• Multimodal Interaction

Innovation

• Dynamic Explainability
• Autoregressive Vision-Language Models
• Attention Head Filtering
• Token-level Attribution
• Visually-grounded Explanation