Dataset · Rapidata
Physics Evaluation in Sora Video Generation
Rapidata/sora-video-generation-physics-likert-scoring
198 generated videos with human physical realism scores for training and evaluating models.
What's inside
Contains 198 video clips, text data, images, and tabular metadata in English in parquet format. For these examples, 5,978 individual ratings were gathered from roughly 6,000 annotators via the Rapidata API to assess the physical plausibility of the generation.
What you can do with it
- Training reward models to evaluate adherence to physical laws in video generators
- Validating text-to-video generation models
- Studying the correlation between human perception of physics and generation metrics
- Tasks
- Video classification, Text → video
- Languages
- English
- Size
- up to 1K rows
- Downloads this month
- 487
- Likes
- 32
Models trained on it
How to load it
from datasets import load_dataset
ds = load_dataset("Rapidata/sora-video-generation-physics-likert-scoring")