Where is the multimodal goal post? On the Ability of Foundation Models to Recognize Contextually Important Moments
Paper
• 2601.16333 • Published
The dataset is currently empty. Upload or create new data files. Then, you will be able to explore them in the Dataset Viewer.
A corpus of video fragments extracted from football games ⚽️ in SoccerReplay-1988. For each fragment the dataset includes an 'importance' annotation (important or non-important), associated audio commentary, and its corresponding textual transcription.
For obtaining access to the MOMENTS dataset, follow these steps:
🔗 More details about the construction and usage of the dataset are available through our preprint:
@misc{surikuchi2026multimodalgoalpostability,
title={Where is the multimodal goal post? On the Ability of Foundation Models to Recognize Contextually Important Moments},
author={Aditya K Surikuchi and Raquel Fernández and Sandro Pezzelle},
year={2026},
eprint={2601.16333},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2601.16333},
}