On Efficient Language and Vision Assistants for Visually-Situated Natural Language Understanding: What Matters in Reading and Reasoning

TL;DR

This paper investigates how to design efficient vision-language models for visually-situated language understanding, focusing on balancing model complexity, computational cost, and performance, and provides practical strategies for optimization.

Contribution

It identifies key components for efficient vision-language models and proposes methods to optimize them, achieving high performance with constrained inference costs.

Findings

Significant improvements in inference throughput without sacrificing accuracy

Insights into the impact of model size and vision modules on performance

Open-source release of models, code, and datasets for reproducibility

Abstract

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.