Offline Preference-Based Reinforcement Learning on Meta-World
63.2Lever Pull Success RateOPRL
Evaluation results
| Method | Links | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| OPRLrandom_seeds=52026.02 | 63.2 | 3.5 | 77.4 | 10.6 | 0.8 | 7.4 | 63.5 | 34.3 | 37.1 | 6.8 | 30.4 | |
| OPRIDErandom_seeds=52026.02 | 51.8 | 79 | 79.9 | 59.1 | 17.7 | 102.2 | 88 | 45.4 | 71.6 | 78.5 | 65.3 | |
| PT+PDSrandom_seeds=52026.02 | 51.7 | 12.4 | 37.3 | 1.8 | 1.1 | 3.4 | 84.3 | 41.5 | 9.2 | 8 | 25 | |
| PTrandom_seeds=52026.02 | 49.2 | 16.8 | 4.9 | 16.7 | 1.1 | 74.8 | 82 | 51.3 | 9.8 | 8 | 31.4 | |
| IDRLrandom_seeds=52026.02 | 33.1 | 67.4 | 79.6 | 30.7 | 14 | 89.2 | 75.8 | 44.3 | 63.1 | 73 | 57 |