Probing Multimodal Large Language Models for Global and Local Semantic Representations

TL;DR

It is found that the intermediate layers of models can encode more global semantic information, whose representation vectors perform better on visual-language entailment tasks, rather than the topmost layers.

Abstract

The advancement of Multimodal Large Language Models (MLLMs) has greatly accelerated the development of applications in understanding integrated texts and images. Recent works leverage image-caption datasets to train MLLMs, achieving state-of-the-art performance on image-to-text tasks. However, there are few studies exploring which layers of MLLMs make the most effort to the global image information, which plays vital roles in multimodal comprehension and generation. In this study, we find that the intermediate layers of models can encode more global semantic information, whose representation vectors perform better on visual-language entailment tasks, rather than the topmost layers. We further probe models regarding local semantic representations through object recognition tasks. We find that the topmost layers may excessively focus on local information, leading to a diminished ability to encode global information. Our code and data are released via https://github.com/kobayashikanna01/probing_MLLM_rep.

Figures (3)

• Figure 1: Performance on the image-text entailment task when using the representations at each layer.
• Figure 2: Performance on the object recognition task when using the representations at each layer of different MLLMs.
• Figure 3: Performance on object recognition of the representations extracted with various prompts.