Atari Games On Atari 57
评估指标
Human World Record Breakthrough
Mean Human Normalized Score
评测结果
各个模型在此基准测试上的表现结果
| Paper Title | Repository | |||
|---|---|---|---|---|
| LBC | 24 | 10077.52% | Learnable Behavior Control: Breaking Atari Human World Records via Sample-Efficient Behavior Selection | - |
| GDI-H3(200M frames) | 22 | 9620.98% | GDI: Rethinking What Makes Reinforcement Learning Different From Supervised Learning | - |
| GDI-H3 | 22 | 9620.33% | Generalized Data Distribution Iteration | - |
| GDI-I3 | 17 | 7810.1% | Generalized Data Distribution Iteration | - |
| MuZero | 19 | 4996.20% | Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model | |
| R2D2 | 15 | 3374.31% | Recurrent Experience Replay in Distributed Reinforcement Learning | - |
| LASER | 7 | 1741.36% | Off-Policy Actor-Critic with Shared Experience Replay | - |
| IMPALA, deep | 3 | 957.34% | IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures | |
| Rainbow DQN | 4 | 873.97% | Rainbow: Combining Improvements in Deep Reinforcement Learning | |
| M-IQN | - | 504% | Munchausen Reinforcement Learning | |
| GDI-H3 | - | - | GDI: Rethinking What Makes Reinforcement Learning Different From Supervised Learning | - |
0 of 11 row(s) selected.