Robot Manipulation on LIBERO (Success Rates and 95% CI)
98.8Spatial Success Rateπ0.5
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| π0.5m=0, s=52026.06 | 98.8 | 98.2 | 98 | 92.4 | 96.9 | — | |
| π0.5 + RTCm=4, s=32026.06 | 98.8 | 99.1 | 97.3 | 92.9 | 97 | — | |
| pi_0.5Additional embodied policy pretraining (P.T.)=Yes2026.06 | 98.8 | 98.2 | 98 | 92.4 | 96.9 | — | |
| π0.5 + RTCm=4, s=22026.06 | 98.7 | 98.3 | 97.4 | 91.5 | 96.5 | — | |
| π0.5 + RTCm=2, s=22026.06 | 98.5 | 97.9 | 97.7 | 92.2 | 96.6 | — | |
| π0.5 + RTCm=3, s=32026.06 | 98.5 | 98.9 | 97.9 | 93.5 | 97.2 | — | |
| LingBot-VAAdditional embodied policy pretraining (P.T.)=Yes2026.06 | 98.5 | 99.6 | 97.2 | 98.5 | 98.5 | — | |
| π0.5 + RTCm=1, s=12026.06 | 98.4 | 94.6 | 95.5 | 90.4 | 94.7 | — | |
| π0.5 + RTCm=4, s=12026.06 | 98.2 | 98.5 | 98.2 | 92.8 | 96.9 | — | |
| Fast-WAMAdditional embodied policy pretraining (P.T.)=No2026.06 | 98.2 | 100 | 97 | 95.2 | 97.6 | — | |
| EinSortBase model=VLA-JEPA, Compression rate=0.82026.06 | 98 | 100 | 98 | 95 | 97.75 | 96.47 | |
| π0 + RTCm=4, s=12026.06 | 97.8 | 99.4 | 97.1 | 84.6 | 94.7 | — | |
| π0-REALFASTm=4, s=22026.06 | 97.7 | 98.9 | 95.7 | 91.5 | 96 | — | |
| π0-REALFASTm=4, s=32026.06 | 97.7 | 98.9 | 95.9 | 90 | 95.6 | — | |
| OriginalBase model=VLA-JEPA, Compression=un-compressed2026.06 | 97.5 | 100 | 99 | 93.5 | 97.5 | 96.17 | |
| π0 + RTCm=1, s=12026.06 | 97.3 | 94.9 | 91.2 | 74.1 | 89.4 | — | |
| π0 + RTCm=2, s=22026.06 | 97.3 | 97.9 | 94.9 | 81.1 | 92.8 | — | |
| π0 + RTCm=3, s=32026.06 | 97.3 | 98.7 | 95 | 83.3 | 93.6 | — | |
| π0 + RTCm=4, s=32026.06 | 97.3 | 98.5 | 96.2 | 87.3 | 94.8 | — | |
| π0 + RTCm=4, s=22026.06 | 97.2 | 98.7 | 96.1 | 85.3 | 94.3 | — | |
| π0-REALFASTm=4, s=42026.06 | 97.2 | 98.6 | 97.1 | 89.9 | 95.7 | — | |
| ImageWAMAdditional embodied policy pretraining (P.T.)=No2026.06 | 97.2 | 99.2 | 98.8 | 98.4 | 98.4 | — | |
| π0m=0, s=52026.06 | 96.8 | 98.8 | 95.8 | 85.2 | 94.2 | — | |
| pi_0Additional embodied policy pretraining (P.T.)=Yes2026.06 | 96.8 | 98.8 | 95.8 | 85.2 | 94.1 | — | |
| MotusAdditional embodied policy pretraining (P.T.)=Yes2026.06 | 96.8 | 99.8 | 96.6 | 97.6 | 97.7 | — | |
| π0-FASTm=0, s=52026.06 | 93 | 94 | 91 | 77 | 88.8 | — | |
| LatentLLMBase model=VLA-JEPA, Compression rate=0.82026.06 | 87 | 64.5 | 59 | 28.5 | 59.75 | 56.31 | |
| OpenVLAAdditional embodied policy pretraining (P.T.)=Yes2026.06 | 84.7 | 88.4 | 79.2 | 53.7 | 76.5 | — | |
| GR00T N1Additional embodied policy pretraining (P.T.)=Yes2026.06 | 84.7 | 88.4 | 79.2 | 53.7 | 76.5 | — | |
| ASVDBase model=VLA-JEPA, Compression rate=0.82026.06 | 62 | 75.5 | 49.5 | 21 | 52 | 48.54 | |
| SVD-LLMBase model=VLA-JEPA, Compression rate=0.82026.06 | 18.5 | 16 | 19 | 0.5 | 13.5 | 11.31 | |
| Plain SVDBase model=VLA-JEPA, Compression rate=0.82026.06 | 0 | 0 | 0 | 0 | 0 | 0 |