Pick-and-place on SIMPLER-WidowX
50.3Average Success Rateπ0-FAST + MG-Select
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| π0-FAST + MG-Selectadditional joint training=true2025.10 | 50.3 | — | — | — | — | 69.4 | 75 | 43.1 | 13.9 | |
| SCALEBackbone=pi0-FAST, Evaluation Protocol=fine-tuned, Training Dataset=BridgeData V22026.02 | 49 | 58.3 | 69.4 | 48.6 | 19.4 | — | — | — | — | |
| π0-FASTreproduced performance=true2025.10 | 46.9 | — | — | — | — | 66.7 | 70.8 | 41.7 | 8.3 | |
| SCALEBackbone=SpatialVLA, Evaluation Protocol=fine-tuned, Training Dataset=BridgeData V22026.02 | 45.1 | 22.2 | 31.9 | 26.4 | 100 | — | — | — | — | |
| SpatialVLA2025.10 | 42.7 | — | — | — | — | 16.7 | 25 | 29.2 | 100 | |
| Top-kBackbone=pi0-FAST, Evaluation Protocol=fine-tuned, k=40, Temperature=0.7, Training Dataset=BridgeData V22026.02 | 41.7 | 48.6 | 61.1 | 41.7 | 15.3 | — | — | — | — | |
| SpatialVLABackbone=SpatialVLA, Evaluation Protocol=fine-tuned, Decoding Strategy=greedy, Training Dataset=BridgeData V22026.02 | 41.7 | 20.8 | 25 | 20.8 | 100 | — | — | — | — | |
| SCALEBackbone=SpatialVLA, Evaluation Protocol=zero-shot, Training Dataset=OXE2026.02 | 41.7 | 22.2 | 34.7 | 34.7 | 75 | — | — | — | — | |
| Top-pBackbone=SpatialVLA, Evaluation Protocol=fine-tuned, p=0.9, Training Dataset=BridgeData V22026.02 | 41.3 | 18.1 | 23.6 | 25 | 98.6 | — | — | — | — | |
| SamplingBackbone=SpatialVLA, Evaluation Protocol=fine-tuned, Temperature=1.0, Training Dataset=BridgeData V22026.02 | 41 | 18.1 | 20.8 | 25 | 100 | — | — | — | — | |
| Top-pBackbone=pi0-FAST, Evaluation Protocol=fine-tuned, p=0.9, Training Dataset=BridgeData V22026.02 | 40.3 | 48.6 | 62.5 | 33.3 | 16.7 | — | — | — | — | |
| Top-kBackbone=SpatialVLA, Evaluation Protocol=fine-tuned, k=40, Temperature=0.7, Training Dataset=BridgeData V22026.02 | 39.2 | 12.5 | 23.6 | 20.8 | 100 | — | — | — | — | |
| SamplingBackbone=pi0-FAST, Evaluation Protocol=fine-tuned, Temperature=1.0, Training Dataset=BridgeData V22026.02 | 35.4 | 41.7 | 54.2 | 29.2 | 16.7 | — | — | — | — | |
| pi0-FASTBackbone=pi0-FAST, Evaluation Protocol=fine-tuned, Decoding Strategy=greedy, Training Dataset=BridgeData V22026.02 | 34.4 | 20.8 | 62.5 | 37.5 | 16.7 | — | — | — | — | |
| SamplingBackbone=SpatialVLA, Evaluation Protocol=zero-shot, Temperature=1.0, Training Dataset=OXE2026.02 | 32.3 | 16.7 | 23.6 | 22.2 | 66.7 | — | — | — | — | |
| SpatialVLABackbone=SpatialVLA, Evaluation Protocol=zero-shot, Decoding Strategy=greedy, Training Dataset=OXE2026.02 | 31.3 | 12.5 | 20.8 | 25 | 66.7 | — | — | — | — | |
| RoboVLM2025.10 | 31.3 | — | — | — | — | 29.2 | 25 | 12.5 | 58.3 | |
| Top-pBackbone=SpatialVLA, Evaluation Protocol=zero-shot, p=0.9, Training Dataset=OXE2026.02 | 30.9 | 13.9 | 26.4 | 22.2 | 61.1 | — | — | — | — | |
| Top-kBackbone=SpatialVLA, Evaluation Protocol=zero-shot, k=40, Temperature=0.7, Training Dataset=OXE2026.02 | 29.9 | 8.3 | 29.2 | 20.8 | 61.1 | — | — | — | — | |
| Octo2025.10 | 16 | — | — | — | — | 12.5 | 8.3 | 0 | 43.1 | |
| RT-1-X2025.10 | 1.1 | — | — | — | — | 0 | 4.2 | 0 | 0 |