Robotic Manipulation on D4RL Adroit pen, door, hammer (human and cloned) v0
52.6Pen Task Score (Human)VIPO-MOBILE
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| VIPO-MOBILErandom seeds=4, dynamics model=value-inconsistency regularized2025.04 | 52.6 | 2 | 1.1 | 71.1 | 2.1 | 25.8 | |
| CQLrandom seeds=42025.04 | 35.2 | 1.2 | 0.6 | 27.2 | 1.4 | 13.1 | |
| MOBILErandom seeds=4, objective=one-step Q(s,a), policy=stochastic, rollouts=short-horizon2025.04 | 30.1 | -0.2 | 0.4 | 69 | 1.5 | 20.2 | |
| BCrandom seeds=42025.04 | 25.8 | 2.8 | 3.1 | 38.3 | 0.7 | 14.1 | |
| MOPOrandom seeds=4, objective=one-step Q(s,a), policy=stochastic2025.04 | 10.7 | -0.2 | 0.3 | 54.6 | 0.5 | 13.2 | |
| TD3+BCrandom seeds=42025.04 | -1 | -0.2 | 0.2 | -2.1 | -0.1 | -0.6 |