Research map / Video / Video generation
Learning Universal Policies via Text-Guided Video Generation
Yilun Du, Sherry Yang, Bo Dai, Hanjun Dai, Ofir Nachum, Joshua B. Tenenbaum, Dale Schuurmans, Pieter Abbeel
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 |
|---|---|---|---|
| condition video model on first observation frame (train-time) (ablation) | UniPi · Combinatorial (seen) Place | accuracy: 13.2 -> 52.4 | Improved |
| condition video model on first observation frame (train-time) (ablation) | UniPi · Combinatorial (seen) Relation | accuracy: 12.4 -> 34.7 | Improved |
| add frame consistency by tiling observed frame (ablation) | UniPi · Combinatorial (seen) Relation | accuracy: 34.7 -> 39.4 | Improved |
| add frame consistency by tiling observed frame (ablation) | UniPi · Combinatorial (seen) Place | accuracy: 52.4 -> 53.2 | No clear effect |
| add temporal hierarchy (temporal super-resolution) (ablation) | UniPi · Combinatorial (seen) Relation | accuracy: 39.4 -> 53.2 | Improved |
| add temporal hierarchy (temporal super-resolution) (ablation) | UniPi · Combinatorial (seen) Place | accuracy: 53.2 -> 59.1 | Improved |
| pretrain on internet-scale video-text data (ablation) | UniPi · Bridge real robot video | FVD: 288.02 -> 264.66 | Improved |
| pretrain on internet-scale video-text data (ablation) | UniPi · Bridge real robot video | FID: 17.75 -> 14.54 | Improved |
| pretrain on internet-scale video-text data (ablation) | UniPi · Bridge real robot video | CLIP score: 24.43 -> 24.54 | No clear effect |
| pretrain on internet-scale video-text data (ablation) | UniPi · Bridge real robot video | success classifier accuracy: 72.6% -> 77.1% | Improved |
| predict future video frames then extract actions instead of diffusing actions directly (Diffuser) | diffusion planner · Combinatorial (novel) Relation | task completion accuracy: 9.6 -> 46.1 | Improved |
Closest papers by meaning
- RoboDreamer: Learning Compositional World Models for Robot Imagination (ICML 2024 · similarity 0.90)
- Video Language Planning (ICLR 2024 · similarity 0.89)
- Solving New Tasks by Adapting Internet Video Knowledge (ICLR 2025 · similarity 0.89)
- Position: Video as the New Language for Real-World Decision Making (ICML 2024 · similarity 0.88)
- Learning to Act from Actionless Videos through Dense Correspondences (ICLR 2024 · similarity 0.88)
- VideoVLA: Video Generators Can Be Generalizable Robot Manipulators (NeurIPS 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.
