Stack on Realman robot collected dataset Real-world In-Domain 1.0 (test)
83.3Success Rate (Cup)CogACT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CogACTPre-training=OXE dataset, Fine-tuning=collected dataset2025.12 | 83.3 | 45.8 | 64.6 | |
| VideoVLAPre-training=OXE dataset, Fine-tuning=collected dataset2025.12 | 75 | 58.3 | 66.7 | |
| π0Pre-training=OXE dataset, Fine-tuning=collected dataset2025.12 | 58.3 | 50 | 54.2 | |
| SpatialVLAPre-training=OXE dataset, Fine-tuning=collected dataset2025.12 | 25 | 16.7 | 20.8 | |
| OpenVLAPre-training=OXE dataset, Fine-tuning=collected dataset2025.12 | 12.5 | 0 | 6.3 |