Put the eggplant into the yellow basket on AutoEval Real-world held-out tasks
78Success RateWorld-Gymnast
Evaluation results
| Method | Links | |
|---|---|---|
| World-GymnastBackbone=600M parameter variant2026.02 | 78 | |
| Iter-SFTBackbone=OpenVLA 7B2026.02 | 17 | |
| SFTBackbone=OpenVLA 7B2026.02 | 8 |