Diffusion-CAM: Faithful Visual Explanations for dMLLMs

arXiv:2604.11005v1 Announce Type: new Abstract: While diffusion Multimodal Large Language Models (dMLLMs) have recently achieved remarkable strides in multimodal generation, the development of interpretability mechanisms has lagged behind their architectural evolution. Unlike traditional autoregressive models that produce sequential activations, diffusion-based architectures generate tokens via parallel denoising, resulting in smooth, distributed activation patterns across the entire sequence. Consequently, existing Class Activation Mapping (CAM) methods, which are tailored for local, sequential dependencies, are ill-suited for interpreting these non-autoregressive behaviors. To bridge this gap, we propose Diffusion-CAM, the first interpretability method specifically tailored for dMLLMs. We derive raw activation maps by differentiably probing intermediate representations in the transformer backbone, accordingly capturing both latent features and their class-specific gradients. To address the inherent stochasticity of these raw signals, we incorporate four key modules to resolve spatial ambiguity and mitigate intra-image confounders and redundant token correlations. Extensive experiments demonstrate that Diffusion-CAM significantly outperforms SoTA methods in both localization accuracy and visual fidelity, establishing a new standard for understanding the parallel generation process of diffusion multimodal systems.

Related

• Pseudo-Unification: Entropy Probing Reveals Divergent Information Patterns in Unified Multimodal Models
• Decoding by Perturbation: Mitigating MLLM Hallucinations via Dynamic Textual Perturbation
• Stabilizing Unsupervised Self-Evolution of MLLMs via Continuous Softened Retracing reSampling
• VL-Calibration: Decoupled Confidence Calibration for Large Vision-Language Models Reasoning
• Faithful-First Reasoning, Planning, and Acting for Multimodal LLMs

Source: arXiv cs.AI | 2026-04-14