Robot Manipulation on Robomimic (Specific Success Rates)
99.5Can Success Rateπ0+FP
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| π0+FPGenerative Mode=Flow Matching, Model Type=VLA & VA2025.10 | 99.5 | 100 | 94 | — | |
| Operational space contr.Safe=false2025.11 | 99 | 100 | 74 | 91 | |
| Florence-Policy-F+FPGenerative Mode=Flow Matching, Model Type=VLA & VA2025.10 | 98.5 | 98.5 | 92.5 | — | |
| π0Generative Mode=Flow Matching, Model Type=VLA2025.10 | 96.5 | 99 | 92.5 | — | |
| Flow Policy (FP)Generative Mode=Flow Matching, Model Type=VA2025.10 | 95 | 13 | 77.5 | — | |
| Florence Policy-FGenerative Mode=Flow Matching, Model Type=VLA2025.10 | 89 | 98.5 | 88.5 | — | |
| PACS - PFLSafe=true2025.11 | 85 | 93 | 38 | 72 | |
| OFFSafe=false2025.11 | 83 | 92 | 34 | 70 | |
| PACS - SSMSafe=true2025.11 | 80 | 97 | 30 | 69 | |
| Florence-Policy-D+MPGenerative Mode=Diffusion, Model Type=VLA & VA2025.10 | 63 | 100 | 54.5 | — | |
| Florence-Policy-D+DPGenerative Mode=Diffusion, Model Type=VLA & VA2025.10 | 62.5 | 100 | 61.5 | — | |
| Florence Policy-DGenerative Mode=Diffusion, Model Type=VLA2025.10 | 61.5 | 97 | 46.5 | — | |
| Diffusion Policy (DP)Generative Mode=Diffusion, Model Type=VA2025.10 | 34.5 | 98.5 | 2 | — | |
| DP+MPGenerative Mode=Diffusion, Model Type=VA & VA2025.10 | 34.5 | 99.5 | 8 | — | |
| Single-action - PFLSafe=true2025.11 | 33 | 94 | 4 | 44 | |
| Single-action - SSMSafe=true2025.11 | 26 | 97 | 0 | 41 | |
| Mamba Policy (MP)Generative Mode=Diffusion, Model Type=VA2025.10 | 5 | 98.5 | 3 | — | |
| Contr. barrier func.Safe=true2025.11 | 0 | 11 | 0 | 4 |