Robot Manipulation on Adroit (Hammer, Door, Pen Scores)
86.7Pen Task ScoreORCA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ORCABackbone=SD 1.52025.10 | 86.7 | — | — | 65.3 | 44 | |
| CLIPType=2D-LT, Input Type=RGB2025.11 | 84 | — | 61 | 72.5 | — | |
| DINOv2Type=2D-LT, Input Type=RGB2025.11 | 84 | — | 76 | 80 | — | |
| AFROType=3D-Pre-D, Input Type=PC2025.11 | 84 | — | 82 | 83 | — | |
| SCRBackbone=SD 1.52025.10 | 84 | — | — | 58 | 32 | |
| CoOpBackbone=SD 1.52025.10 | 82.7 | — | — | 58 | 33.3 | |
| TADPBackbone=SD 1.52025.10 | 81.3 | — | — | 57.3 | 33.3 | |
| DP3Type=3D-Pol, Input Type=PC2025.11 | 80 | — | 70 | 75 | — | |
| Drift-Based PolicyNFE=12026.04 | 80 | 100 | 70 | — | — | |
| Text (Simple)Backbone=SD 1.52025.10 | 80 | — | — | 57.3 | 34.7 | |
| Text (Caption)Backbone=SD 1.52025.10 | 80 | — | — | 57.3 | 34.7 | |
| PointDifType=3D-Pre-S, Input Type=PC2025.11 | 78 | — | 76 | 77 | — | |
| PointMAEType=3D-Pre-S, Input Type=PC2025.11 | 76 | — | 64 | 70 | — | |
| DynaMo-3DType=3D-Pre-D, Input Type=PC2025.11 | 76 | — | 73 | 74.5 | — | |
| FVPType=3D-Pre-D, Input Type=PC2025.11 | 76 | — | 66 | 71 | — | |
| ReV2026.04 | 73 | — | 79 | — | — | |
| PointNetType=3D-LT, Input Type=PC2025.11 | 72 | — | 80 | 76 | — | |
| DynaMoType=2D-Pre-D, Input Type=RGB2025.11 | 68 | — | 76 | 72 | — | |
| CDP2026.04 | 68 | — | 74 | — | — | |
| VC-1Backbone=ViT-L/162025.10 | 65.3 | — | — | 47.3 | 29.3 | |
| OMP-DDENFE (Number of Function Evaluations)=12025.12 | 64 | 100 | 68 | 80.8 | — | |
| Ada3DriftNFE=1, Input Modality=3D point cloud-based2026.03 | 63.3 | 90.3 | 65 | — | — | |
| Simple DP3NFE=10, Input Modality=3D point cloud-based2026.03 | 61 | 86 | 61 | — | — | |
| OMP-JVPNFE (Number of Function Evaluations)=12025.12 | 60 | 100 | 68 | 82.3 | — | |
| OMPNFE=12026.04 | 60 | 100 | 68 | — | — | |
| DP3NFE=10, Input Modality=3D point cloud-based2026.03 | 59.7 | 88.7 | 64.2 | — | — | |
| CLIPBackbone=ViT-L/162025.10 | 58.7 | — | — | 51.4 | 44 | |
| MP1NFE (Number of Function Evaluations)=12025.12 | 58 | 100 | 69 | 78.9 | — | |
| OMP-JVP w/o L_disNFE (Number of Function Evaluations)=12025.12 | 58 | 100 | 69 | 81.2 | — | |
| FlowPolicyNFE=1, Input Modality=3D point cloud-based2026.03 | 58 | 77 | 61.2 | — | — | |
| MP1NFE=12026.04 | 58 | 100 | 69 | — | — | |
| MP1NFE=1, Input Modality=3D point cloud-based2026.03 | 57.7 | 84.3 | 64.2 | — | — | |
| FlowPolicyNFE (Number of Function Evaluations)=12025.12 | 54 | 98 | 61 | 71.6 | — | |
| FlowPolicyNFE=12026.04 | 54 | 98 | 61 | — | — | |
| DP32026.04 | 53 | — | 69 | — | — | |
| OMP-JVP w/o L_dis & L_DANFE (Number of Function Evaluations)=12025.12 | 48 | 95 | 66 | 78.3 | — | |
| ACT2026.04 | 47 | — | 66 | — | — | |
| DP3NFE (Number of Function Evaluations)=102025.12 | 46 | 100 | 56 | 68.7 | — | |
| DP3NFE=102026.04 | 46 | 100 | 56 | — | — | |
| Simple DP3NFE (Number of Function Evaluations)=102025.12 | 36 | 98 | 40 | 67.4 | — | |
| Simple DP3NFE=102026.04 | 36 | 98 | 40 | — | — | |
| Adaflow2025.12 | 18 | 45 | 27 | 35.6 | — | |
| AdaFlowInput Modality=image-based2026.03 | 18 | 45 | 27 | — | — | |
| AdaflowNFE=—2026.04 | 18 | 45 | 27 | — | — | |
| DPNFE (Number of Function Evaluations)=102025.12 | 13 | 16 | 34 | 35.2 | — | |
| CPNFE (Number of Function Evaluations)=12025.12 | 13 | 45 | 31 | 50.1 | — | |
| DPNFE=10, Input Modality=image-based2026.03 | 13 | 16 | 34 | — | — | |
| CPNFE=1, Input Modality=image-based2026.03 | 13 | 45 | 31 | — | — | |
| DPNFE=102026.04 | 13 | 16 | 34 | — | — | |
| CPNFE=12026.04 | 13 | 45 | 31 | — | — |