Visual Planning on VP2 benchmark
68Push Success Rate (Robosuite)DiLA
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| DiLAaverage of independent runs=4, normalization=relative to the ground truth simulator baseline, protocol=visual planning protocol2026.05 | 68 | 15 | 78.33 | 35.83 | 20.83 | 3.33 | 41.44 | |
| ADAWORLDaverage of independent runs=4, normalization=relative to the ground truth simulator baseline, protocol=visual planning protocol2026.05 | 63.5 | 5.83 | 29.17 | 10.83 | 10 | 5 | 21.54 |