Intuitive physics benchmark for video and VLM models
A large English WorldBench benchmark for evaluating physical understanding in generative and vision models.
- Visual question answering
- Text → video
- English
Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.
Answered by Gemini 4 · the chat is kept only in this tab
A large English WorldBench benchmark for evaluating physical understanding in generative and vision models.
Unedited daily and working first-person video recordings in English for training video models.
Over 8,100 English and Chinese prompts together with generated videos and preview frames.
English image and data collection from the VideoGrain paper (ICLR 2025) for multi-level video editing tasks.
Large multi-domain image-text corpus for 4D world modeling and robotics.
44,482 examples and over 780,000 human ratings to test LLM capability in creating 3D objects via code.
400,476 videos with previews, view counts, explicit flags, and captions for video-text models.
1 to 10 million before-and-after video pairs with English text instructions for 10 editing tasks.
1,313 examples with 74.8 thousand human ratings for videos generated by the Marey model.
Dataset of up to 1 thousand examples for evaluating image-to-video generation and digital human animation.
A benchmark for evaluating goal-directed manipulations in video generation using egocentric frames.
198 generated videos with human physical realism scores for training and evaluating models.
A large dataset of ~70K full HD circular car video reviews for video generation tasks.