Vision-Language-Action on LIBERO
99.4Success Rate (Spatial)VLA-ADP
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| VLA-ADPCKPT=OFT (7B), Ratio=50%2026.03 | 99.4 | 98 | 96.4 | 91.2 | 96.3 | 1.23 | |
| TeamVLACKPT=OFT (7B)2026.03 | 99.2 | 96.5 | 97 | 93.8 | 96.6 | 1.51 | |
| VLA-ADPCKPT=OFT (7B), Ratio=70%2026.03 | 99 | 98.2 | 96.8 | 91.2 | 96.3 | 1.13 | |
| OpenVLA-OFTCKPT=OFT (7B)2026.03 | 98.6 | 98.2 | 96.6 | 94.8 | 97.1 | 1 | |
| VLA-CacheCKPT=OFT (7B)2026.03 | 98.3 | 97.5 | 98.3 | 95.4 | 97.4 | 1.3 | |
| SpecPrune-VLACKPT=OFT (7B)2026.03 | 98.2 | 96.3 | 97.7 | 94 | 96.6 | 1.46 | |
| VLA-ADPCKPT=OFT (7B), Ratio=30%2026.03 | 97.6 | 98.4 | 97.4 | 84.2 | 94.4 | 1.35 | |
| VLA-IAPCKPT=OFT (7B), Ratio=70%2026.03 | 97.6 | 99.6 | 98.4 | 95.6 | 97.8 | 1.25 | |
| VLA-IAPCKPT=OFT (7B), Ratio=50%2026.03 | 97.3 | 99.1 | 98.2 | 95.2 | 97.5 | 1.37 | |
| CogACTCKPT=CogVLM (7B)2026.03 | 97.2 | 98 | 90.2 | 88.8 | 93.6 | — | |
| FastVCKPT=OFT (7B)2026.03 | 96.8 | 81 | 96.4 | 73 | 86.8 | 1.24 | |
| VLA-IAPCKPT=OFT (7B), Ratio=30%2026.03 | 96.6 | 98.8 | 98 | 94.8 | 97.1 | 1.54 | |
| EfficientVLACKPT=OFT (7B)2026.03 | 96.5 | 91.1 | 96 | 72.1 | 88.9 | 1.54 | |
| SmolVLACKPT=SmolVLM (2.2B)2026.03 | 93 | 94 | 91 | 77 | 88.8 | — | |
| SemiVLA + Selective LoRASup.=10%, Trainable Params=31.6M, Adaptation=Semi-supervised2026.06 | 92.4 | 89.5 | 91.8 | 82.3 | 89 | — | |
| SemiVLA + LoRASup.=10%, Trainable Params=79.4M, Adaptation=Semi-supervised2026.06 | 91.7 | 88.6 | 90.9 | 80.5 | 87.9 | — | |
| SemiVLA + QLoRASup.=10%, Trainable Params=79.4M, Adaptation=Semi-supervised2026.06 | 90.8 | 87.4 | 89.7 | 78.9 | 86.7 | — | |
| SemiVLA + Full FTSup.=10%, Trainable Params=7.0B, Adaptation=Semi-supervised2026.06 | 90.5 | 87.8 | 89.2 | 78.5 | 86.5 | — | |
| SemiVLA + AdapterSup.=10%, Trainable Params=42.7M, Adaptation=Semi-supervised2026.06 | 88.9 | 84.7 | 87.2 | 74.6 | 83.9 | — | |
| WorldVLACKPT=Cham. (7B)2026.03 | 87.6 | 96.2 | 83.4 | 60 | 81.8 | — | |
| OpenVLA + LoRASup.=10%, Trainable Params=79.4M, Adaptation=Supervised2026.06 | 86.8 | 82.1 | 84.9 | 70.3 | 81 | — | |
| OpenVLA + QLoRASup.=10%, Trainable Params=79.4M, Adaptation=Supervised2026.06 | 85.9 | 80.5 | 83.7 | 68.9 | 79.8 | — | |
| NORACKPT=Qwen-VL (3B)2026.03 | 85.6 | 87.8 | 77 | 45 | 73.9 | — | |
| OpenVLACKPT=OpenVLA (7B)2026.03 | 84.7 | 88.4 | 79.2 | 53.7 | 76.5 | — | |
| OpenVLA + AdapterSup.=10%, Trainable Params=42.7M, Adaptation=Supervised2026.06 | 84.6 | 78.2 | 81.5 | 65.7 | 77.5 | — | |
| OpenVLA + Full FTSup.=10%, Trainable Params=7.0B, Adaptation=Supervised2026.06 | 84.5 | 79.8 | 82.6 | 66.8 | 78.4 | — | |
| OpenVLA + Linear ProbeSup.=10%, Trainable Params=15.2M, Adaptation=Supervised2026.06 | 81.3 | 72.4 | 76.8 | 59.6 | 72.5 | — | |
| OpenVLASup.=0, Trainable Params=0, Adaptation=Zero-shot2026.06 | 80 | 69.6 | 74 | 55.5 | 69.8 | — |