This research introduces a novel benchmark dataset specifically designed for evaluating how multimodal large language models (MLLMs) interpret eye examination images.

• Combines fundus photographs and OCT images with detailed clinical metadata
• Tests AI models on their ability to diagnose common eye conditions like diabetic retinopathy
• Evaluates performance across multiple state-of-the-art MLLMs including GPT-4V and Gemini Pro
• Identifies current limitations in medical visual reasoning for ophthalmic applications

This benchmark addresses a critical gap in MLLM evaluation for specialized medical domains, potentially accelerating the development of AI assistants for ophthalmologists and improving diagnostic accuracy.

A Novel Ophthalmic Benchmark for Evaluating Multimodal Large Language Models with Fundus Photographs and OCT Images