Robot Manipulation on ALOHA (in-domain)
96.2Pick and Place Success RateQwen-VLA-alohaw/ pretrain
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Qwen-VLA-alohaw/ pretrainPre-training strategy=fine-tuned from Qwen-VLA-Base2026.05 | 96.2 | 92.3 | 98.7 | 87.2 | 65.4 | 61.5 | 83.6 | |
| π0.52026.05 | 73.1 | 84.6 | 88.5 | 69.2 | 80.8 | 33.3 | 71.6 | |
| GR00T N1.62026.05 | 30.8 | 38.5 | 53.8 | 19.2 | 19.2 | 10.3 | 28.6 | |
| Qwen-VLA-alohaw/o pretrainPre-training strategy=trained from scratch2026.05 | 30.8 | 53.8 | 61.5 | 64.1 | 50 | 30.8 | 48.5 |