Robotic Manipulation on Semantic OOD
70Inverse SuccessNoTVLA
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| NoTVLATraining Steps=5k2025.10 | 70 | 69 | 62 | 78 | 69.8 | |
| NoTVLATraining Steps=30k2025.10 | 63 | 62 | 49 | 84 | 64.5 | |
| π0.5Training Steps=5k2025.10 | 1 | 58 | 59 | 62 | 45 | |
| Qwen-PITraining Steps=30k2025.10 | 0 | 35 | 23 | 44 | 25.5 | |
| Qwen-PITraining Steps=5k2025.10 | 0 | 23 | 45 | 59 | 31.8 | |
| π0Training Steps=30k2025.10 | 0 | 30 | 25 | 32 | 21.8 | |
| π0Training Steps=5k2025.10 | 0 | 18 | 56 | 55 | 32.3 | |
| π0.5Training Steps=30k2025.10 | 0 | 48 | 60 | 57 | 41.3 |