Goal-conditioned visual planning on CrossTask T=3 88
47.47Success Rate (SR)GeoWorld
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GeoWorldBackbone=ViT-g3842026.02 | 47.47 | 73.69 | 86.55 | |
| GeoWorldBackbone=ViT-g2026.02 | 46.25 | 71.95 | 85.44 | |
| V-JEPA 2Backbone=ViT-g3842026.02 | 45.58 | 72.74 | 69.42 | |
| GeoWorldBackbone=ViT-H2026.02 | 45.33 | 70.84 | 84.7 | |
| V-JEPA 2Backbone=ViT-g2026.02 | 44.84 | 71.62 | 68.87 | |
| V-JEPA 2Backbone=ViT-H2026.02 | 44.07 | 70.18 | 68.32 | |
| GeoWorldBackbone=ViT-L2026.02 | 43.89 | 68.96 | 82.93 | |
| V-JEPA 2Backbone=ViT-L2026.02 | 43.33 | 68.63 | 67.84 | |
| MTID2026.02 | 40.45 | 67.19 | 69.17 | |
| SCHEMA2026.02 | 38.93 | 63.8 | 79.82 | |
| KEPP2026.02 | 38.12 | 64.74 | 67.15 | |
| ActionDiffusion2026.02 | 37.79 | 65.38 | 67.45 | |
| PDPP2026.02 | 37.2 | 64.67 | 66.57 | |
| LFP2026.02 | 30.55 | 59.59 | 76.86 | |
| VidAssistzero-shot=false2026.02 | 28.85 | 58.12 | 75.36 | |
| E3P2026.02 | 26.4 | 53.02 | 74.05 | |
| P3IV2026.02 | 23.34 | 49.96 | 73.89 | |
| Ext-MGAIL2026.02 | 21.27 | 49.46 | 61.7 | |
| Int-MGAIL2026.02 | 17.03 | 44.66 | 58.08 | |
| PlaTe2026.02 | 16 | 36.17 | 65.91 | |
| VidAssistzero-shot=true2026.02 | 14.6 | 52.6 | 68.38 | |
| DDN2026.02 | 12.18 | 31.29 | 47.48 | |
| Retrieval-Based2026.02 | 8.05 | 23.3 | 32.06 | |
| UPN2026.02 | 2.89 | 24.39 | 31.56 | |
| UAAA2026.02 | 2.15 | 20.21 | 30.87 | |
| WLTDO2026.02 | 1.87 | 21.64 | 31.7 | |
| Random2026.02 | 0.01 | 0.94 | 1.66 |