Procedure Planning on CrossTask long horizon (T=3 to T=6)
37.96Success Rate (T=3)OursHow
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OursHowFeature=HowTo100M2023.03 | 37.96 | 22.56 | 14.3 | 8.93 | |
| OursBaseFeature=Base, Supervision=C2023.03 | 26.43 | 16.2 | 9.2 | 6.28 | |
| P3IVFeature=HowTo100M, Supervision=L2023.03 | 23.34 | 13.4 | 7.21 | 4.4 | |
| DDNFeature=Base, Supervision=V2023.03 | 12.18 | 5.97 | 3.1 | 1.2 | |
| Retrieval-BasedFeature=Base2023.03 | 8.05 | 3.95 | 2.4 | 1.1 |