NingLab/EcomMMMU
Introduction 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… See the full description on the dataset page: https://huggingface.co/datasets/NingLab/EcomMMMU.
Introduction
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 understanding. It includes a specialized vision-salient subset (VSS), designed to test scenarios where textual information alone is insufficient and visuals are crucial.
Dataset Sources
- Repository: GitHub
<!-- ## Data Split The statistics for the MMECInstruct Dataset are shown in the table below.
-->
Quick Start
Run the following command to get the data:
from datasets import load_dataset
dataset = load_dataset("NingLab/EcomMMMU")License
Please check the license of each subset in our curated dataset ECInstruct.
Citation
@article{ling2025ecommmmu,
title={EcomMMMU: Strategic Utilization of Visuals for Robust Multimodal E-Commerce Models},
author={Ling, Xinyi and Du, Hanwen and Zhu, Zhihui and Ning, Xia},
journal={arXiv preprint arXiv:2508.15721},
year={2025}
}