Explaining Multi-modal Large Language Models by Analyzing their Vision Perception
This addresses the problem of interpretability for users in critical applications, offering a novel method for analyzing vision perception in MLLMs.
The paper tackles the interpretability challenge in Multi-modal Large Language Models (MLLMs) by proposing a novel architecture that integrates an open-world localization model with a MLLM, enabling simultaneous text generation and object localization from vision embeddings, which enhances interpretability through saliency maps, hallucination identification, and bias assessment.
Multi-modal Large Language Models (MLLMs) have demonstrated remarkable capabilities in understanding and generating content across various modalities, such as images and text. However, their interpretability remains a challenge, hindering their adoption in critical applications. This research proposes a novel approach to enhance the interpretability of MLLMs by focusing on the image embedding component. We combine an open-world localization model with a MLLM, thus creating a new architecture able to simultaneously produce text and object localization outputs from the same vision embedding. The proposed architecture greatly promotes interpretability, enabling us to design a novel saliency map to explain any output token, to identify model hallucinations, and to assess model biases through semantic adversarial perturbations.