Research map / Video / Video segmentation and tracking
Dyn-VPP: Video Prediction Policy Optimization for Improved Visual Dynamics
Zirui Ge, Pengxiang Ding, Baohua Yin, Yemin Wang, Qishen Wang, Zhiyong Xie, Hengtao Li, Runze Suo, Wenxuan Song, Han Zhao, Shangke Lyu, Haoang Li and 5 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 |
|---|---|---|---|
| add GRPO post-training of the video prediction model, AGM frozen (ablation) | VPP base policy · CALVIN ABC->D | average trajectory length: 4.28 -> 4.35 | Improved |
| add GRPO post-training of VPM, then retrain the action generation model (ablation) | VPP base policy · CALVIN ABC->D | average trajectory length: 4.28 -> 4.56 | Improved |
| retrain the AGM after post-training the VPM (versus frozen AGM) (ablation) | Dyn-VPP · CALVIN ABC->D | average trajectory length: 4.35 -> 4.56 | Improved |
| add post-training of VPM, AGM frozen (ablation) | VPP base policy · CALVIN ABC->D | 5 tasks in a row (%): 74.7 -> 77.2 | Improved |
| add post-training of VPM and AGM (ablation) | VPP base policy · CALVIN ABC->D | 5 tasks in a row (%): 74.7 -> 83.1 | Improved |
| add Dyn-VPP post-training (ablation) | VPP base policy · CALVIN ABC->D | average effective rank: 29.28 -> 43.88 | Improved |
| add Dyn-VPP post-training (ablation) | VPP base policy · CALVIN ABC->D | average effective rank ratio: 0.0603 -> 0.0814 | Improved |
| add Dyn-VPP post-training | VPP · L-CALVIN ABC->D | average trajectory length: 5.53 -> 6.73 | Improved |
| add Dyn-VPP post-training | VPP · L-CALVIN ABC->D | 10 tasks in sequence: 0.28 -> 0.39 | Improved |
| use Dyn-VPP instead of Tri-VLA extension of VPP | VPP · CALVIN ABC->D | average trajectory length: 4.37 -> 4.56 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Vidman — VidMan: Exploiting Implicit Dynamics from Video Diffusion Model for Effective Robot Manipulation (NeurIPS 2024) · better on 1 of 1 shared result
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
- Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representations (ICML 2025 · similarity 0.90)
- VideoVLA: Video Generators Can Be Generalizable Robot Manipulators (NeurIPS 2025 · similarity 0.89)
- DiT4DiT: Jointly Modeling Video Dynamics and Actions for Generalizable Robot Control (NeurIPS 2026 · similarity 0.88)
- VidMan: Exploiting Implicit Dynamics from Video Diffusion Model for Effective Robot Manipulation (NeurIPS 2024 · similarity 0.88)
- ElegantVLA: Learning When to Think for Efficient Vision-Language-Action Models (NeurIPS 2026 · similarity 0.87)
- Prediction with Action: Visual Policy Learning via Joint Denoising Process (NeurIPS 2024 · 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.
