Robot Manipulation on RoboVerse
96CloseBox Success RateVGDP
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| VGDPEncoder Architecture=Visual-Geometry Diffusion Policy (VGDP), Input Modality=Visual-Geometry, Policy Head=Diffusion-policy, Randomization Level Averaging=True2025.11 | 96 | 3 | 79 | 88 | 53 | 72 | 65 | |
| PointCloud_spUnetEncoder Architecture=SparseUNet, Input Modality=PointCloud, Policy Head=Diffusion-policy, Randomization Level Averaging=True2025.11 | 75 | 0 | 0 | 0 | 8 | 0 | 14 | |
| PointCloud_DP3Encoder Architecture=DP3, Input Modality=PointCloud, Policy Head=Diffusion-policy, Randomization Level Averaging=True2025.11 | 58 | 0 | 0 | 0 | 7 | 0 | 11 | |
| RGBD_resnet18Encoder Architecture=ResNet18, Input Modality=RGBD, Policy Head=Diffusion-policy, Randomization Level Averaging=True2025.11 | 53 | 6 | 24 | 62 | 18 | 11 | 29 | |
| RGBD_MultiViTEncoder Architecture=MultiViT, Input Modality=RGBD, Policy Head=Diffusion-policy, Randomization Level Averaging=True2025.11 | 49 | 1 | 48 | 69 | 33 | 14 | 33 | |
| RGBD_ViTEncoder Architecture=ViT, Input Modality=RGBD, Policy Head=Diffusion-policy, Randomization Level Averaging=True2025.11 | 48 | 0 | 61 | 68 | 31 | 13 | 32 | |
| RGB_resnet18Encoder Architecture=ResNet18, Input Modality=RGB, Policy Head=Diffusion-policy, Randomization Level Averaging=True2025.11 | 35 | 10 | 22 | 72 | 17 | 8 | 27 |