PhysLoop leverages large language models to guide video generation, selection, and correction, enabling physically consistent self-improvement and controllable data construction for video models.
Video generation models produce visually realistic videos but still struggle with physical dynamics and causal consistency. Learning from generated videos requires fine-grained physical supervision, yet reliable quality judgments, correction pairs, and controlled variations are difficult to obtain. Large language models (LLMs) possess physical knowledge, but struggle to reliably guide video generation models in producing their training videos. We present PhysLoop, a unified harness designed to support physically grounded self-improvement in video generation models. The harness guides and constrains LLMs throughout video generation, selection, and correction, distilling their physical…
video generation modelsphysical dynamicscausal consistencylarge language modelsphysically grounded self-improvementcontrollable training-data constructioninference-time video repairrecursive self-improvementcontrollable generationlocalized repair
- Upvotes
- 11
- Authors
- 14
- Published
- Nov 12, 2026
Summary and abstract excerpt via Hugging Face Papers ↗ · arXiv.