On Efficient Language and Vision Assistants for Visually-Situated Natural Language Understanding: What Matters in Reading and Reasoning
This work addresses the challenge of computational efficiency for researchers and practitioners in vision-language AI, though it appears incremental as it builds on existing open-source models.
The study tackled the problem of high computational demands in complex visually-situated text understanding by redefining vision-language model design to achieve efficient models with constrained inference costs, resulting in significant improvements in inference throughput while maintaining high performance across models from 160M to 13B parameters.
Recent advancements in language and vision assistants have showcased impressive capabilities but suffer from a lack of transparency, limiting broader research and reproducibility. While open-source models handle general image tasks effectively, they face challenges with the high computational demands of complex visually-situated text understanding. Such tasks often require increased token inputs and large vision modules to harness high-resolution information. Striking a balance between model size and data importance remains an open question. This study aims to redefine the design of vision-language models by identifying key components and creating efficient models with constrained inference costs. By strategically formulating datasets, optimizing vision modules, and enhancing supervision techniques, we achieve significant improvements in inference throughput while maintaining high performance. Extensive experiments across models ranging from 160M to 13B parameters offer insights into model optimization. We will fully open-source our codebase, models, and datasets at https://github.com/naver-ai/elva.