Drifting Field Policy: A One-Step Generative Policy via Wasserstein Gradient Flow
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866910201349668864 |
|---|---|
| author | Koo, Juil Park, Mingue Choi, Jiwon Min, Yunhong Sung, Minhyuk |
| author_facet | Koo, Juil Park, Mingue Choi, Jiwon Min, Yunhong Sung, Minhyuk |
| contents | We propose Drifting Field Policy (DFP), a non-ODE one-step generative policy built on the drifting model paradigm. We frame the policy update as a reverse-KL Wasserstein-2 gradient flow toward a soft target policy, so that each DFP update corresponds to a gradient step in probability space. By construction, this gradient is decomposed into an ascent toward higher action-value regions and a score matching with the anchor policy as a trust region. We further derive a simple, tractable surrogate of the otherwise intractable update loss, akin to behavior cloning on top-K critic-selected actions. We find empirically that this mechanism uniquely benefits the drifting backbone owing to its non-ODE parameterization. With one-step inference, DFP achieves state-of-the-art performance on several manipulation tasks across Robomimic and OGBench, outperforming ODE-based policies. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_07727 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Drifting Field Policy: A One-Step Generative Policy via Wasserstein Gradient Flow Koo, Juil Park, Mingue Choi, Jiwon Min, Yunhong Sung, Minhyuk Machine Learning Artificial Intelligence Robotics We propose Drifting Field Policy (DFP), a non-ODE one-step generative policy built on the drifting model paradigm. We frame the policy update as a reverse-KL Wasserstein-2 gradient flow toward a soft target policy, so that each DFP update corresponds to a gradient step in probability space. By construction, this gradient is decomposed into an ascent toward higher action-value regions and a score matching with the anchor policy as a trust region. We further derive a simple, tractable surrogate of the otherwise intractable update loss, akin to behavior cloning on top-K critic-selected actions. We find empirically that this mechanism uniquely benefits the drifting backbone owing to its non-ODE parameterization. With one-step inference, DFP achieves state-of-the-art performance on several manipulation tasks across Robomimic and OGBench, outperforming ODE-based policies. |
| title | Drifting Field Policy: A One-Step Generative Policy via Wasserstein Gradient Flow |
| topic | Machine Learning Artificial Intelligence Robotics |
| url | https://arxiv.org/abs/2605.07727 |