NU-World-Model-Embodied-AI/phyground
Dataset · https://huggingface.co/datasets/NU-World-Model-Embodied-AI/phyground
PhyGround: Benchmarking Physical Reasoning in Generative World Models Project page · Paper · Evaluation code · PhyJudge-9B PhyGround is a criteria-grounded benchmark for diagnosing physical failures in generated video. It contains 250 prompts covering 13 observable physical laws across solid-body mechanics, fluid dynamics, and optics. Each prompt is paired with a first-frame image, 10 released generation configurations, and applicable-law labels. The Hub repository includes:… See the full description on the dataset page: https://huggingface.co/datasets/NU-World-Model-Embodied-AI/phyground.
First seen by Oriel on .
AI summary
PhyGround is a benchmark for diagnosing physical reasoning failures in generated video using 250 prompts, 13 physical laws, model outputs, and quality-controlled human annotations.
- The dataset covers solid-body mechanics, fluid dynamics, and optics, with each prompt labeled for two to four applicable physical laws.
- It includes 2,500 generated videos across 10 configuration directories, while released human annotations cover the paper's 2,000-video, 8-model benchmark set.
- Videos are scored for semantic alignment, physical temporal validity, object persistence, and applicable physical laws on a 1–5 Likert scale.
- The repository provides prompts, conditioning images, videos, annotation files, schemas, and download commands for reproducible evaluation.
- The benchmark does not model exhaustive physics, validate hidden causal state, certify safety, or support generalizing rankings to later model versions without retesting.
#video generation #physical reasoning #world models #benchmarking
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Datasets: 40, latest observation ; +17 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Datasets: 3,285, latest observation ; +1,969 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Datasets | NU-World-Model-Embodied-AI/phyground | https://huggingface.co/datasets/NU-World-Model-Embodied-AI/phyground Read discussion |