MultiInteract-Bench is a comprehensive dataset designed to evaluate the capabilities of multimodal large language models in reproducing web-based interactions from image sequences. The dataset contains real-world web interface snapshots showing progressive states of web applications through user interactions.

	Key Features… See the full description on the dataset page: https://huggingface.co/datasets/zionzionzion/MultiInteract-Bench.