SportR: A Benchmark for Multimodal Large Language Model Reasoning in Sports
Paper • 2511.06499 • Published
Datasets:
Request access to the SportR media
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
The SportR media (images and video clips) are sourced from publicly available broadcast / online footage and are provided for non-commercial academic research only. The QA annotations, Chain-of-Thought rationales and bounding boxes are the authors' original work. Access is granted on a case-by-case basis. By requesting access you agree to all the terms below.
Log in or Sign Up to review the conditions and access this dataset content.
SportR
SportR is a multi-sport benchmark for rule-infraction reasoning and tactic recognition across basketball, soccer, American football, badminton, and table tennis, covering both images (4,789) and video clips (2,052), with a progressive QA suite and 6,841 human-authored Chain-of-Thought rationales.
⚠️ Access policy. The annotations / QA JSON are open (see the GitHub repository). The image & video media in this repository are gated and released for non-commercial research only — they are derived from third-party broadcast/online footage whose copyright belongs to the original owners. If you are a rights holder and wish to have specific content removed, please contact haotian.xia@rice.edu and we will promptly remove it.
Contents
SportsImage/— 4,789 images ({sport}_{N}.{ext}, e.g.basketball_1.png,american_football_5.png,soccer_123.png).SportsVideo/— 2,052 video clips ({Sport}{N}.mp4, e.g.Basketball122.mp4).- QA / CoT annotations (JSON) are hosted openly on
GitHub; each record's
pathfield points to a file underSportsImage/orSportsVideo/.
Citation
@misc{xia2026sportrbenchmarkmultimodallarge,
title={SportR: A Benchmark for Multimodal Large Language Model Reasoning in Sports},
author={Haotian Xia and Haonan Ge and Junbo Zou and Hyun Woo Choi and Xuebin Zhang and Danny Suradja and Botao Rui and Ethan Tran and Wendy Jin and Zhen Ye and Xiyang Lin and Christopher Lai and Shengjie Zhang and Junwen Miao and Shichao Chen and Rhys Tracy and Vicente Ordonez and Weining Shen and Hanjie Chen},
year={2026},
eprint={2511.06499},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2511.06499},
}
License
Annotations: CC BY-NC 4.0. Media: research-use-only under the access agreement above; copyrights remain with the original footage owners.
- Downloads last month
- 55