Rethinking VLMs and LLMs for Image Classification

• Published in: arXiv.org
• Main Authors: Cooper, Avi, Kato, Keizo, Shih, Chia-Hsien, Yamane, Hiroaki, Vinken, Kasper, Takemoto, Kentaro, Sunagawa, Taro, Hao-Wei Yeh, Yamanaka, Jin, Mason, Ian, Boix, Xavier
• Format: Paper
• Language: English
• Published: Ithaca Cornell University Library, arXiv.org 03.10.2024
• Subjects: Accuracy; Cost effectiveness; Datasets; Image classification; Image enhancement; Large language models; Lightweight; Object recognition; Performance enhancement; Visual tasks; Weight reduction

Visual Language Models (VLMs) are now increasingly being merged with Large Language Models (LLMs) to enable new capabilities, particularly in terms of improved interactivity and open-ended responsiveness. While these are remarkable capabilities, the contribution of LLMs to enhancing the longstanding key problem of classifying an image among a set of choices remains unclear. Through extensive experiments involving seven models, ten visual understanding datasets, and multiple prompt variations per dataset, we find that, for object and scene recognition, VLMs that do not leverage LLMs can achieve better performance than VLMs that do. Yet at the same time, leveraging LLMs can improve performance on tasks requiring reasoning and outside knowledge. In response to these challenges, we propose a pragmatic solution: a lightweight fix involving a relatively small LLM that efficiently routes visual tasks to the most suitable model for the task. The LLM router undergoes training using a dataset constructed from more than 2.5 million examples of pairs of visual task and model accuracy. Our results reveal that this lightweight fix surpasses or matches the accuracy of state-of-the-art alternatives, including GPT-4V and HuggingGPT, while improving cost-effectiveness.