Interaction-limited Policy Learning on Procgen (easy mode)
36.3Bigfish ScoreExpert Videos
Evaluation Results
| Method | Links | ||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Expert Videostype=oracle/baseline2025.12 | 36.3 | 10 | 9.7 | 9.9 | 11.5 | 13.5 | 8.5 | 10.2 | 29.9 | 67 | 9.5 | 11.9 | 9.2 | 7.4 | 10 | 11.6 | 16.6 | 100 | |
| BCV-LRExpert Videos=true, Environmental Rewards=false, Interaction limit=100k2025.12 | 35.9 | 9.9 | 9.3 | 8.9 | 4.4 | 12.4 | 7.5 | 9.4 | 27.5 | 54.8 | 7.2 | 11.6 | 4.6 | 4 | 3.1 | 10.3 | 13.8 | 79 | |
| UPESVExpert Videos=true, Environmental Rewards=false, Interaction limit=100k2025.12 | 30.5 | 9.7 | 9.4 | 7.4 | 3.5 | 9.1 | 6.6 | 6.8 | 20.6 | 15 | 6.3 | 9.3 | 3.5 | 2.9 | 0.8 | 2 | 9 | 58 | |
| LAPOExpert Videos=true, Environmental Rewards=true, Interaction limit=100k2025.12 | 20.6 | 9.6 | 9.4 | 6.2 | 4.8 | 5.9 | 7.3 | 4.7 | 0.5 | 20.3 | 5.2 | 6.7 | 3.9 | 2.7 | 0.8 | 0.3 | 6.8 | 48 | |
| BCOExpert Videos=true, Environmental Rewards=false, Interaction limit=100k2025.12 | 3.6 | 7.4 | 7.6 | 6.7 | 4.2 | 5.4 | 6.4 | 3.3 | 3.5 | 12.8 | 4.2 | 5.8 | 2.8 | 2.5 | 0.8 | 0.4 | 4.8 | 38 | |
| PPOExpert Videos=false, Environmental Rewards=true, Interaction limit=100k2025.12 | 0.9 | 5 | 3.7 | 4.1 | 4.4 | 1.1 | 3.5 | 2.2 | -1.9 | 2.6 | 3.4 | 1.2 | 3 | 2.6 | 0.4 | 0.1 | 2.3 | 22 | |
| ILPOExpert Videos=true, Environmental Rewards=false, Interaction limit=100k2025.12 | 0.8 | 4.2 | 6.7 | 3.7 | 2.2 | 0.6 | 3.1 | 3.4 | -2 | 0.5 | 2.2 | 1.2 | 3.2 | 2.6 | 0.7 | 0.1 | 2.1 | 22 | |
| LAIFOExpert Videos=true, Environmental Rewards=false, Interaction limit=100k2025.12 | 0.8 | 4.3 | 5.4 | 4.7 | 3.2 | 0.8 | 3.9 | 2.7 | -2.5 | 2 | 3 | 1.2 | 2.4 | 1.9 | 0.6 | 0.3 | 2.2 | 22 |