Research map / Video / Action recognition
DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos
Shenyuan Gao, William Liang, Kaiyuan Zheng, Ayaan Naveed Malik, Seonghyeon Ye, Sihyun Yu, Wei-Cheng Tseng, Yuzhu Dong, Kaichun Mo, Chen-Hsuan Lin, Jiannan Xiang, Yuqi Xie and 11 more
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 |
|---|---|---|---|
| DreamDojo (ablation) | In-lab Eval · pretrain with continuous latent actions instead of action-free video | PSNR: 20.797 -> 20.913 | Improved |
| DreamDojo (ablation) | EgoDex Eval · pretrain with continuous latent actions instead of action-free video | PSNR: 19.924 -> 20.344 | Improved |
| Cosmos-Predict2.5 (ablation) | In-lab / EgoDex Eval · action-free pretraining on human videos vs no pretraining | PSNR: 20.576 -> 20.797 (In-lab); 19.952 -> 19.924 (EgoDex) | No clear effect |
| DreamDojo (ablation) | In-lab Eval · use ground-truth actions instead of latent actions | PSNR: 20.913 -> 20.960 | Improved |
| DreamDojo (ablation) | In-lab Eval · add EgoDex to In-lab pretraining mixture | PSNR: 20.913 -> 20.972 | Improved |
| DreamDojo (ablation) | DreamDojo-HV Eval · add DreamDojo-HV to pretraining mixture | PSNR: 18.706 -> 18.724 | Improved |
| DreamDojo | Counterfactual Eval · scale model from 2B to 14B | PSNR: 20.907 -> 21.087 | Improved |
| Cosmos-Predict2.5 (ablation) | Counterfactual Eval · pretrain on human videos vs none | PSNR: 20.472 -> 20.907 | Improved |
| DreamDojo-2B | GR-1 Long Eval · distill teacher into few-step autoregressive student | PSNR: 14.086 -> 13.146 | Mixed |
| DreamDojo-2B | GR-1 Long Eval · distill teacher into few-step autoregressive student | FPS: 2.72 -> 10.93 | Improved |
| GR00T N1.5 policy | AgiBot fruit packing · apply model-based planning with DreamDojo vs uniform sampling | success rate: 55.33% -> 68.00% | Improved |
Closest papers by meaning
- DreamVLA: A Vision-Language-Action Model Dreamed with Comprehensive World Knowledge (NeurIPS 2025 · similarity 0.89)
- EA-WM: Event-Aware Generative World Model with Structured Kinematic-to-Visual Action Fields (NeurIPS 2026 · similarity 0.89)
- Predictive Inverse Dynamics Models are Scalable Learners for Robotic Manipulation (ICLR 2025 · similarity 0.89)
- Exploring the Limits of Vision-Language-Action Manipulation in Cross-task Generalization (NeurIPS 2025 · similarity 0.88)
- Robots Pre-train Robots: Manipulation-Centric Robotic Representation from Large-Scale Robot Datasets (ICLR 2025 · similarity 0.88)
- VideoVLA: Video Generators Can Be Generalizable Robot Manipulators (NeurIPS 2025 · similarity 0.88)
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.
