Robotic Manipulation on RoboTwin 2.0 (Short/Long Horizon Success Rates)
86Short Horizon Success RateCortex
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Cortexdata-scaling setting=50 clean + 500 randomized demonstrations2026.07 | 86 | 88 | 86.8 | |
| π0.5data-scaling setting=50 clean + 500 randomized demonstrations2026.07 | 82.6 | 82.95 | 82.74 | |
| X-VLAdata-scaling setting=50 clean + 500 randomized demonstrations2026.07 | 77.13 | 66.3 | 72.8 | |
| DP3data-scaling setting=50 clean + 500 randomized demonstrations2026.07 | 62.67 | 53.45 | 55.24 | |
| π0data-scaling setting=50 clean + 500 randomized demonstrations2026.07 | 61.5 | 72.55 | 65.92 | |
| OpenVLA-OFTdata-scaling setting=50 clean + 500 randomized demonstrations2026.07 | 42.1 | 32.8 | 38.3 | |
| RDT-1Bdata-scaling setting=50 clean + 500 randomized demonstrations2026.07 | 35.73 | 32.75 | 34.5 | |
| ACTdata-scaling setting=50 clean + 500 randomized demonstrations2026.07 | 33.2 | 24.5 | 29.7 |