Alibaba's PhysOPSD uses executable code as privileged context to make video generation physically plausible
AlibabaTongyiLab · hf · 2026-11-11
HKUST and Alibaba Token Hub propose PhysOPSD, improving the physical plausibility of video generative models via on-policy self-distillation.
- Core idea: use executable code as privileged physical context to inject real physical constraints into generation
- Trained with dual-level privilege and controlled variations
- Project page shows clearly improved physical plausibility over the base model in scenes like falling dominoes and pouring water
Code and project page are available.
More from Multimodal
- Opus 5.5 generates a restrained, watchable line-art animation of streets and rivers — FinanceYF5 · 2026-11-11
- AI-generated holographic cyberpunk shop scene impresses with detail — Zard-Oz · 2026-11-11
- Seedream 5.5 edits hold character identity across 30 frames in user test — aziz4ai · 2026-11-11
- Tsinghua and Shengshu's APEX replaces fake-score nets with condition shifting for faster streaming video generation — Tsinghua · 2026-11-11
- AI short film 'A Single Fry' shows cinematic seagull scene — gps2019 · 2026-11-11
- Designer: fable 5.2 plus paper is an unstoppable combo for extending existing visual languages — evilrabbit_ · 2026-11-11