What Large Language Models Bring to Text-rich VQA?
This work addresses text-rich VQA, a cross-modal task important for applications like document analysis, by showing that LLMs can enhance performance, though it is incremental as it builds on existing OCR and LLM methods.
The paper investigates the use of Large Language Models (LLMs) for text-rich Visual Question Answering (VQA), finding that a training-free pipeline combining external OCR with LLMs achieves superior performance on four datasets compared to most existing Multimodal LLMs, with LLMs providing stronger comprehension and helpful knowledge.
Text-rich VQA, namely Visual Question Answering based on text recognition in the images, is a cross-modal task that requires both image comprehension and text recognition. In this work, we focus on investigating the advantages and bottlenecks of LLM-based approaches in addressing this problem. To address the above concern, we separate the vision and language modules, where we leverage external OCR models to recognize texts in the image and Large Language Models (LLMs) to answer the question given texts. The whole framework is training-free benefiting from the in-context ability of LLMs. This pipeline achieved superior performance compared to the majority of existing Multimodal Large Language Models (MLLM) on four text-rich VQA datasets. Besides, based on the ablation study, we find that LLM brings stronger comprehension ability and may introduce helpful knowledge for the VQA problem. The bottleneck for LLM to address text-rich VQA problems may primarily lie in visual part. We also combine the OCR module with MLLMs and pleasantly find that the combination of OCR module with MLLM also works. It's worth noting that not all MLLMs can comprehend the OCR information, which provides insights into how to train an MLLM that preserves the abilities of LLM.