EcomMMMU is a large-scale multimodal multitask understanding dataset for e-commerce applications,
containing 406,190 samples and 8,989,510 product images across 34 product categories.
It is designed to systematically evaluate how multimodal large language models (MLLMs)
utilize visual information in real-world shopping scenarios.
Unlike prior datasets that treat all images equally,
EcomMMMU explicitly investigates when and how multiple product images contribute to… See the full description on the dataset page:
https://huggingface.co/datasets/NingLab/EcomMMMU.
