Bimanual Multi-Task Learning on RoboTwin and RLBench average over all tasks 2
1,200NpRDT
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| RDTAveraged over all tasks=true2026.01 | 1,200 | — | 1,200 | 183.1 | |
| SMPAveraged over all tasks=true2026.01 | 258.9 | — | 80.2 | 107.3 | |
| Discrete PolicyAveraged over all tasks=true2026.01 | 162.9 | — | 162.9 | 153.7 | |
| Sparse DPAveraged over all tasks=true2026.01 | 154.4 | — | 110.1 | 148.3 | |
| DPAveraged over all tasks=true2026.01 | 132.5 | — | 132.5 | 120.3 | |
| DP3Averaged over all tasks=true2026.01 | 128.5 | — | 128.5 | 122.1 | |
| ACTAveraged over all tasks=true2026.01 | 83.9 | — | 83.9 | 94.8 |