Research map / Video / Video generation
PhysPO: Physics-Aware Local Preference Optimization for Physically Consistent Video Diffusion
Zhuoran Yang, Yanyong Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace random dispreferred videos with CounterPhy counterfactual pairs (ablation) | PhysPO (Wan2.1-1.3B) · PhyGenBench | PC: 38.4 -> 46.0 | Improved |
| replace random dispreferred videos with CounterPhy counterfactual pairs (ablation) | PhysPO (Wan2.1-1.3B) · PhyGenBench | SA: 35.0 -> 43.0 | Improved |
| use PA-DPO + Tie-DPO instead of vanilla DPO on CounterPhy (ablation) | PhysPO (Wan2.1-1.3B) · PhyGenBench | PC: 26.0 -> 46.0 | Improved |
| add masked physics-aware PA-DPO term (ablation) | PhysPO (Wan2.1-1.3B) · PhyGenBench | PC: 29.4 -> 46.0 | Improved |
| add tie-aware neutrality regularization Tie-DPO (ablation) | PhysPO (Wan2.1-1.3B) · PhyGenBench | SA: 34.6 -> 43.0 | Improved |
| add tie-aware neutrality regularization Tie-DPO (ablation) | PhysPO (Wan2.1-1.3B) · PhyGenBench | PC: 40.2 -> 46.0 | Improved |
| apply PhysPO post-training to CogVideoX-5B | CogVideoX-5B · VideoPhy (overall) | PC: 31.4 -> 49.4 | Improved |
| apply PhysPO post-training to Wan2.1-T2V-1.3B | Wan2.1-T2V-1.3B · VideoPhy (overall) | PC: 24.0 -> 42.6 | Improved |
| apply PhysPO post-training to Wan2.1-T2V-1.3B | Wan2.1-T2V-1.3B · PhyGenBench | SA: 30.0 -> 43.0 | Improved |
| PhysPO vs PhyGDPO and VideoDPO on same backbone | Wan2.1-T2V-1.3B · VideoPhy2 overall | overall score: 0.1407 -> 0.1470 (vs PhyGDPO) | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- CogVideoX-5B — CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer (ICLR 2025) · better on 4 of 4 shared results
- PhyGDPO — PhyGDPO: Physics-Aware Groupwise Direct Preference Optimization for Physically Consistent Text-to-Video Generation (ECCV 2026) · better on 2 of 2 shared results
- WISA — WISA: World simulator assistant for physics-aware text-to-video generation (NeurIPS 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PhysVid: Physics Aware Local Conditioning for Generative Video Models (CVPR 2026 · similarity 0.93)
- Hierarchical Fine-grained Preference Optimization for Physically Plausible Video Generation (NeurIPS 2025 · similarity 0.91)
- PhysCtrl: Generative Physics for Controllable and Physics-Grounded Video Generation (NeurIPS 2025 · similarity 0.88)
- LaMo: Self-Supervised Latent Motion Priors for Physical Realism in Video Generation (NeurIPS 2026 · similarity 0.87)
- DenseDPO: Fine-Grained Temporal Preference Optimization for Video Diffusion Models (NeurIPS 2025 · similarity 0.87)
- VideoDPO: Omni-Preference Alignment for Video Diffusion Generation (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
