Robot Manipulation on RoboTwin Hard 2.0 (Task Success Rates and 50-Task Average)
45Click Alarm Success RateKAM-WM
Evaluation Results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| KAM-WMDemonstrations=50, Rollouts per task=1002026.07 | 45 | 9 | 25 | 40 | 69 | 80 | 35 | 31 | 59 | 43.7 | 22.4 | |
| DP3Demonstrations=50, Rollouts per task=1002026.07 | 14 | 1 | 3 | 18 | 7 | 3 | 0 | 3 | 19 | 7.6 | 5 | |
| RDTDemonstrations=50, Rollouts per task=1002026.07 | 12 | 16 | 5 | 27 | 32 | 31 | 14 | 24 | 45 | 22.9 | 13.7 | |
| π0 (FT)Demonstrations=50, Rollouts per task=100, Protocol=Fine-tuning2026.07 | 11 | 3 | 2 | 4 | 46 | 13 | 8 | 29 | 60 | 19.6 | 16.3 | |
| DPDemonstrations=50, Rollouts per task=1002026.07 | 5 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 8 | 1.4 | 0.6 | |
| ACTDemonstrations=50, Rollouts per task=1002026.07 | 4 | 0 | 0 | 0 | 0 | 0 | 0 | 6 | 10 | 2.2 | 1.7 |