Capturing Visual Environment Structure Correlates with Control Performance
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_ | 1866908813601275904 |
|---|---|
| author | Dong, Jiahua Man, Yunze Tokmakov, Pavel Wang, Yu-Xiong |
| author_facet | Dong, Jiahua Man, Yunze Tokmakov, Pavel Wang, Yu-Xiong |
| contents | The choice of visual representation is key to scaling generalist robot policies. However, direct evaluation via policy rollouts is expensive, even in simulation. Existing proxy metrics focus on the representation's capacity to capture narrow aspects of the visual world, like object shape, limiting generalization across environments. In this paper, we take an analytical perspective: we probe pretrained visual encoders by measuring how well they support decoding of environment state -- including geometry, object structure, and physical attributes -- from images. Leveraging simulation environments with access to ground-truth state, we show that this probing accuracy strongly correlates with downstream policy performance across diverse environments and learning settings, significantly outperforming prior metrics and enabling efficient representation selection. More broadly, our study provides insight into the representational properties that support generalizable manipulation, suggesting that learning to encode the latent physical state of the environment is a promising objective for control. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2602_04880 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Capturing Visual Environment Structure Correlates with Control Performance Dong, Jiahua Man, Yunze Tokmakov, Pavel Wang, Yu-Xiong Robotics The choice of visual representation is key to scaling generalist robot policies. However, direct evaluation via policy rollouts is expensive, even in simulation. Existing proxy metrics focus on the representation's capacity to capture narrow aspects of the visual world, like object shape, limiting generalization across environments. In this paper, we take an analytical perspective: we probe pretrained visual encoders by measuring how well they support decoding of environment state -- including geometry, object structure, and physical attributes -- from images. Leveraging simulation environments with access to ground-truth state, we show that this probing accuracy strongly correlates with downstream policy performance across diverse environments and learning settings, significantly outperforming prior metrics and enabling efficient representation selection. More broadly, our study provides insight into the representational properties that support generalizable manipulation, suggesting that learning to encode the latent physical state of the environment is a promising objective for control. |
| title | Capturing Visual Environment Structure Correlates with Control Performance |
| topic | Robotics |
| url | https://arxiv.org/abs/2602.04880 |