Language-conditioned robot manipulation on LIBERO-Goal Clean
98Success RateStableVLA
Evaluation Results
| Method | Links | |
|---|---|---|
| StableVLATraining Method=VLM Direct FT, Parameters=0.5B2026.05 | 98 | |
| VLA-AdapterTraining Method=VLM Direct FT, Parameters=0.5B2026.05 | 97.4 | |
| OpenPi–0.5Training Method=OpenX+Web Co-train, Parameters=3B2026.05 | 97.2 | |
| OpenVLA-OFTTraining Method=OpenX Pretrain, Parameters=7B2026.05 | 96.8 | |
| OpenVLATraining Method=OpenX Pretrain, Parameters=7B2026.05 | 74 |