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.
Large collection of video clips with English captions for training video generation models.
106 hours of free-camera flights across 17 games for 3D reconstruction and video generation.
About 123,000 fully annotated synthetic video clips covering warehouse safety scenarios.
10,000 hours of real human work video with head and hand pose annotations in video format for robotics.
English image and data collection from the VideoGrain paper (ICLR 2025) for multi-level video editing tasks.
About 400k English video clips with detailed scene, shot, and camera motion descriptions.
Large multi-domain image-text corpus for 4D world modeling and robotics.
4,320 audio-visual videos from 15 accelerated versions of the MiniMax-H3 model for evaluating generation quality in few steps.
Repackaged OpenVid-1M dataset into .tar shards for fast streaming during video model training.
A collection of 14M videos with text descriptions for training text-to-video diffusion models.
Dataset for the Vera model (Netflix and Caltech) for layered video editing with background preservation.
Million-scale dataset of text prompts and generated videos for text-to-video research.
7.8k videos as latent arrays and CLIP prompt embeddings for video generation.
10 thousand five-second Street Fighter III videos with player action tables and text descriptions.
Part of the training sample with video and 3D data for training models to understand physics.
A shard of a multimodal dataset containing video and 3D data for training models to understand physical laws in visual environments.
Part (shard 05/16) of the PhysInOne training dataset featuring video and 3D data for physical reasoning tasks.
Millions of videos and 3D assets for training models to understand physical laws and generate video.
Training part (6 of 16) of the PhysInOne dataset for video generation and understanding physical laws.
A set of high-quality 4K/8K videos with structured captions for video generation.
142,220 human motion animations with text descriptions and annotations for humanoid robotics.