Continual Imitation Learning on LIBERO goal 1.0
87.9FWT (Forward Transfer)UniVLA
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| UniVLAModel Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 87.9 | 57.7 | 36.5 | 29.6 | |
| UniVLAModel Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 87.5 | 65.7 | 35.2 | 20.5 | |
| TS-StellarModel Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 84.8 | 37.2 | 50.8 | 41 | |
| π0.5Model Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 84.5 | 38.4 | 55.4 | 43.8 | |
| π0Model Category=VLA Models, Training Strategy=Scratch2025.11 | 82.9 | 49 | 45.5 | 35.7 | |
| TS-StellarModel Category=CIL Methods, Training Strategy=Pretrained, Pretrained Data Source=LIBERO-902025.11 | 82.7 | 19.4 | 64.8 | 57.3 | |
| T-StellarModel Category=VLA Models, Training Strategy=Scratch2025.11 | 81.4 | 20.7 | 61.7 | 67.9 | |
| T-StellarModel Category=CIL Methods, Training Strategy=Scratch2025.11 | 81.4 | 20.7 | 61.7 | 67.9 | |
| T-StellarModel Category=CIL Methods, Training Strategy=Pretrained, Pretrained Data Source=LIBERO-902025.11 | 80.4 | 12.6 | 66.7 | 62.1 | |
| ERModel Category=CIL Methods, Training Strategy=Pretrained, Pretrained Data Source=LIBERO-902025.11 | 80.3 | 21.1 | 60.5 | 55.3 | |
| SeqLoRAModel Category=CIL Methods, Training Strategy=Pretrained, Pretrained Data Source=LIBERO-902025.11 | 80.3 | 78.8 | 26.2 | 9.5 | |
| TS-StellarModel Category=VLA Models, Training Strategy=Scratch2025.11 | 80.2 | 22.2 | 60.7 | 64.2 | |
| TS-StellarModel Category=CIL Methods, Training Strategy=Scratch2025.11 | 80.2 | 22.2 | 60.7 | 64.2 | |
| T-StellarModel Category=VLA Models, Training Strategy=Scratch2025.11 | 79.6 | 31 | 49.7 | 42.9 | |
| T-StellarModel Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 76.6 | 19.1 | 57.5 | 57 | |
| π0.5Model Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 75 | 41.7 | 37.9 | 35.3 | |
| π0Model Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 74.9 | 25.6 | 55.6 | 41.1 | |
| TS-StellarModel Category=VLA Models, Training Strategy=Scratch2025.11 | 73.3 | 27.8 | 48.3 | 42.6 | |
| π0Model Category=VLA Models, Training Strategy=Scratch2025.11 | 72.7 | 46.9 | 31.1 | 24.9 | |
| TS-StellarModel Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 71.8 | 25.9 | 47.5 | 47.7 | |
| MoDEModel Category=VLA Models, Training Strategy=Scratch2025.11 | 71 | 36.2 | 39.7 | 33.1 | |
| ERModel Category=CIL Methods, Training Strategy=Scratch2025.11 | 70.8 | 33 | 41.8 | 34.2 | |
| π0Model Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 70.3 | 33.5 | 40.8 | 34.5 | |
| MoDEModel Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 69.7 | 23.2 | 46.8 | 41.6 | |
| IsCiLModel Category=CIL Methods, Training Strategy=Pretrained, Pretrained Data Source=LIBERO-902025.11 | 66.8 | 7.8 | 60.9 | 48.5 | |
| MoDEModel Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 52.9 | 11.9 | 43 | 40.3 | |
| MoDEModel Category=VLA Models, Training Strategy=Scratch2025.11 | 48.2 | 15.4 | 33.6 | 28.5 | |
| LoTUSModel Category=CIL Methods, Training Strategy=Pretrained, Pretrained Data Source=LIBERO-902025.11 | 46.8 | 37.7 | 18.5 | 10.8 | |
| LoTUSModel Category=CIL Methods, Training Strategy=Scratch2025.11 | 44.8 | 31.3 | 22.2 | 16.3 | |
| T-StellarModel Category=VLA Models, Training Strategy=Pretrained, Pretrained Data Source=cross-embodiment data2025.11 | 43.8 | 4.1 | 38.3 | 35.8 | |
| UniVLAModel Category=VLA Models, Training Strategy=Scratch2025.11 | 35.5 | 28.1 | 9.4 | 12 | |
| SeqLoRAModel Category=CIL Methods, Training Strategy=Scratch2025.11 | 26.2 | 23.9 | 8.9 | 2.4 | |
| IsCiLModel Category=CIL Methods, Training Strategy=Scratch2025.11 | 24.3 | 3 | 22 | 18.5 | |
| UniVLAModel Category=VLA Models, Training Strategy=Scratch2025.11 | 3.1 | 3.1 | 0.4 | 0 |