VBVR-Dataset: Very Big Video Reasoning Training Data
- Type
- dataset
- Venue
- Video-Reason
- Year
- 2026
- Source
- huggingface
- Access
- free
- Language
- en
- Added
- 2026-07-17T20:00:41.318240+00:00
- Verified
- 2026-07-17T20:00:41.318240+00:00
Summary
This dataset is designed to support large-scale training and scaling studies of reasoning capabilities in video generation models.
Keywords
hf-dataset qa---multimodal parquet text datasets pandas polars mlcroissant video-reasoning video-generation visual-reasoning benchmark spatiotemporal vbvr has-paper
Topics
QA / Multimodal
Research notes
- downloads=4264; likes=55