Reinforcement Learning on Atari 26 Game Subset
47.5Median ScoreAPT
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| APTRL interaction steps=100K, Contrastive Learning (CL)=true2021.03 | 47.5 | 69.55 | 7 | |
| APTRL interaction steps=100K, Contrastive Learning (CL)=false2021.03 | 21.23 | 28.12 | 3 | |
| VISRRL interaction steps=100K2021.03 | 9.5 | 128.07 | 7 | |
| CBBRL interaction steps=100K, Contrastive Learning (CL)=true2021.03 | 1.78 | 17.34 | 2 | |
| DIAYNRL interaction steps=100K, Contrastive Learning (CL)=true2021.03 | 1.76 | 28.44 | 2 | |
| DIAYNRL interaction steps=100K, Contrastive Learning (CL)=false2021.03 | 1.34 | 25.39 | 2 | |
| CBBRL interaction steps=100K, Contrastive Learning (CL)=false2021.03 | 1.23 | 21.94 | 3 | |
| MEPOLRL interaction steps=100K, Contrastive Learning (CL)=true2021.03 | 1.05 | 21.78 | 3 | |
| MEPOLRL interaction steps=100K, Contrastive Learning (CL)=false2021.03 | 0.34 | 17.94 | 2 |