Browse and Concentrate: Comprehending Multimodal Content via prior-LLM Context Fusion
This addresses a specific bottleneck in MLLMs for multi-image understanding, offering incremental improvements in domain-specific tasks.
The paper tackles the problem of multimodal large language models (MLLMs) struggling to comprehend context involving multiple images due to prior-LLM modality isolation, and proposes a browse-and-concentrate paradigm that boosts average accuracy by 2.13% and 7.60% against strong baselines with 3B and 11B LLMs, respectively, across 7 multi-image scenarios.
With the bloom of Large Language Models (LLMs), Multimodal Large Language Models (MLLMs) that incorporate LLMs with pre-trained vision models have recently demonstrated impressive performance across diverse vision-language tasks. However, they fall short to comprehend context involving multiple images. A primary reason for this shortcoming is that the visual features for each images are encoded individually by frozen encoders before feeding into the LLM backbone, lacking awareness of other images and the multimodal instructions. We term this issue as prior-LLM modality isolation and propose a two phase paradigm, browse-and-concentrate, to enable in-depth multimodal context fusion prior to feeding the features into LLMs. This paradigm initially "browses" through the inputs for essential insights, and then revisits the inputs to "concentrate" on crucial details, guided by these insights, to achieve a more comprehensive understanding of the multimodal inputs. Additionally, we develop training strategies specifically to enhance the understanding of multi-image inputs. Our method markedly boosts the performance on 7 multi-image scenarios, contributing to increments on average accuracy by 2.13% and 7.60% against strong MLLMs baselines with 3B and 11B LLMs, respectively.