---
license: apache-2.0
library_name: custom
language:
- en
pipeline_tag: video-to-video
base_model: Wan-AI/Wan2.2-I2V-A14B
datasets:
- NU-World-Model-Embodied-AI/phyground
tags:
- world-model
- video-generation
- physical-ai
- physics
- flow-matching
- dpo
- wan2.2
- arxiv:2605.19242
---

# PhyWorld: Physics-Faithful Video World Model

[Project page](https://nu-world-model-embodied-ai.github.io/PhyWorld/) ·
[Paper](https://arxiv.org/abs/2605.19242) ·
[Code and release status](https://github.com/NU-World-Model-Embodied-AI/PhyWorld) ·
[PhyGround benchmark](https://huggingface.co/datasets/NU-World-Model-Embodied-AI/phyground)

PhyWorld is a physics-faithful video world model post-trained from
[Wan2.2-I2V-A14B](https://huggingface.co/Wan-AI/Wan2.2-I2V-A14B). It is
designed to continue an observed scene while improving temporal coherence and
physical plausibility.

The model is trained in two stages:

1. **Video-continuation flow matching** to learn coherent continuation from an
   observed video prefix.
2. **Physics-preference DPO** using preference pairs derived from the
   [PhyGround](https://phyground.github.io/) human-annotation pool.

> **Usage status:** this repository contains the released checkpoint, but it is
> not a generic `DiffusionPipeline` and the public code repository does not yet
> contain a verified end-to-end PhyWorld inference command. Download and file
> inspection are supported below. Follow the
> [GitHub release status](https://github.com/NU-World-Model-Embodied-AI/PhyWorld)
> for the official inference entry point rather than substituting a Wan2.2 or
> generic Diffusers command.

## Model details

| Field | Value |
| --- | --- |
| Base model | [Wan2.2-I2V-A14B](https://huggingface.co/Wan-AI/Wan2.2-I2V-A14B) |
| Primary task | Video-to-video scene continuation |
| Post-training | Continuation flow matching, then physics-preference DPO |
| Preference source | [PhyGround](https://huggingface.co/datasets/NU-World-Model-Embodied-AI/phyground) |
| Training hardware reported in the paper | 16 × NVIDIA H100 |
| Languages | English prompts were used in the reported evaluation |
| License | Apache-2.0; see the license section and base-model terms |

## Download

Install the current Hugging Face CLI and download the snapshot:

```bash
pip install -U huggingface_hub

hf download NU-World-Model-Embodied-AI/phyworld \
  --local-dir ./phyworld
```

The complete snapshot is approximately 70 GB. To inspect only the model card
and configs before downloading weights:

```bash
hf download NU-World-Model-Embodied-AI/phyworld \
  README.md configuration.json \
  high_noise_model/config.json low_noise_model/config.json \
  --local-dir ./phyworld-config
```

## Checkpoint layout

```text
phyworld/
├── high_noise_model/
│   ├── config.json
│   └── diffusion.safetensors
├── low_noise_model/
│   ├── config.json
│   └── diffusion.safetensors
├── Wan2.1_VAE.pth
├── models_t5_umt5-xxl-enc-bf16.pth
├── google/umt5-xxl/
└── configuration.json
```

The two denoiser directories follow the high-noise/low-noise expert layout
used by Wan2.2. The snapshot also includes the VAE and UMT5 text-encoder
artifacts required by the released checkpoint. File names are kept as released
for compatibility.

## Reported results

### PhyGround physical faithfulness

Scores use the released PhyJudge-9B evaluation protocol on a 1–5 scale.

| Model | SA | PTV | Persistence | Solid-body | Fluid | Optical | Overall |
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| **PhyWorld** | **2.78** | **3.07** | **3.23** | **2.84** | **3.04** | **3.57** | **3.09** |
| Wan2.2-I2V-A14B | 2.72 | 2.97 | 3.08 | 2.79 | 3.03 | 3.36 | 2.99 |

### VBench general video quality

| Model | Subject consistency | Background consistency | Motion smoothness | Dynamic degree | Aesthetic quality | Imaging quality | Average |
| --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| **PhyWorld** | **0.932** | **0.944** | **0.986** | **0.564** | **0.555** | **0.632** | **0.769** |
| Wan2.2-I2V-A14B | 0.912 | 0.928 | 0.977 | 0.554 | 0.543 | 0.622 | 0.756 |

These are the authors' reported results under the paper's protocol. They are
not a guarantee of performance on different prompts, resolutions, inference
implementations, or downstream domains.

## Training data

Stage 1 uses OpenVid-1M clips filtered by inter-frame CLIP similarity and
optical-flow magnitude. Stage 2 uses 1,000 class-balanced preference pairs
derived from PhyGround human annotations across seven event classes. See the
paper for filtering thresholds, DPO settings, and ablations.

## Intended use

PhyWorld is intended for:

- research on physically faithful video generation and world models;
- controlled comparison with Wan2.2 and other video-generation baselines;
- evaluation on PhyGround and general video-quality benchmarks;
- further research on physics-aware post-training.

This is a research checkpoint, not a validated physics simulator, safety model,
or robot controller. Generated video should not be treated as evidence that an
event is physically possible or as a substitute for simulation or real-world
testing.

## Limitations

- PhyWorld can still produce physically implausible motion, object changes, and
  long-horizon inconsistencies.
- The reported physical improvement is benchmark- and prompt-dependent.
- PhyGround covers 13 observable laws across solid-body mechanics, fluids, and
  optics; it does not cover all physical phenomena.
- The public repository does not yet provide a verified end-to-end inference
  script, a minimum-memory configuration, or a complete reproducibility recipe.
- Biases and coverage limitations in Wan2.2, OpenVid-1M, and the preference data
  may be inherited by this checkpoint.

## License

This checkpoint is released under the Apache License 2.0. PhyWorld is derived
from [Wan2.2-I2V-A14B](https://huggingface.co/Wan-AI/Wan2.2-I2V-A14B), which is
also released under Apache-2.0. Users remain responsible for reviewing all
applicable upstream terms and the terms of any data or software used with the
model.

## Citation

```bibtex
@misc{zhao2026phyworld,
  title         = {PhyWorld: Physics-Faithful World Model for Video Generation},
  author        = {Pu Zhao and Juyi Lin and Timothy Rupprecht and Arash Akbari and Chence Yang and Rahul Chowdhury and Elaheh Motamedi and Arman Akbari and Yumei He and Chen Wang and Geng Yuan and Weiwei Chen and Yanzhi Wang},
  year          = {2026},
  eprint        = {2605.19242},
  archivePrefix = {arXiv},
  primaryClass  = {cs.CV},
  url           = {https://arxiv.org/abs/2605.19242}
}
```