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.
086
1---2license: cc-by-4.03task_categories:4- text-classification5language:6- en7tags:8- e-commerce9size_categories:10- 100K<n<1M11---12 13## Introduction14EcomMMMU is a large-scale multimodal multitask understanding dataset for e-commerce applications, 15containing 406,190 samples and 8,989,510 product images across 34 product categories. 16It is designed to systematically evaluate how multimodal large language models (MLLMs) 17utilize visual information in real-world shopping scenarios.18 19Unlike prior datasets that treat all images equally, 20EcomMMMU explicitly investigates when and how multiple product images contribute to understanding. 21It includes a specialized vision-salient subset (VSS), 22designed to test scenarios where textual information alone is insufficient and visuals are crucial.23 24 25## Dataset Sources26- **Repository:** [GitHub](https://github.com/ninglab/EcomMMMU)27 28<!-- ## Data Split29The statistics for the MMECInstruct Dataset are shown in the table below.30 31| Split | Size |32| --- | --- |33| Train | 56,000 |34| Validation | 7,000 |35 -->36 37## Quick Start38 39Run the following command to get the data:40 41```python42from datasets import load_dataset43 44dataset = load_dataset("NingLab/EcomMMMU")45```46 47## License48Please check the license of each subset in our curated dataset ECInstruct.49 50| Dataset | License Type |51| --- | --- |52| [Amazon Review](https://amazon-reviews-2023.github.io/) | Non listed |53| [AmazonQA](https://github.com/amazonqa/amazonqa) | Non listed |54| [Shopping Queries Dataset](https://github.com/amazon-science/esci-data) | Apache License 2.0 |55 56## Citation57```bibtex58@article{ling2025ecommmmu,59 title={EcomMMMU: Strategic Utilization of Visuals for Robust Multimodal E-Commerce Models},60 author={Ling, Xinyi and Du, Hanwen and Zhu, Zhihui and Ning, Xia},61 journal={arXiv preprint arXiv:2508.15721},62 year={2025}63}64```