Metric: Human World Record Breakthrough (higher is better)
| # | Model↕ | Human World Record Breakthrough▼ | Extra Data | Paper | Date↕ | Code |
|---|---|---|---|---|---|---|
| 1 | LBC | 24 | No | Learnable Behavior Control: Breaking Atari Human... | 2023-05-09 | - |
| 2 | GDI-H3(200M frames) | 22 | No | GDI: Rethinking What Makes Reinforcement Learnin... | 2021-06-11 | - |
| 3 | GDI-H3 | 22 | No | Generalized Data Distribution Iteration | 2022-06-07 | - |
| 4 | MuZero | 19 | No | Mastering Atari, Go, Chess and Shogi by Planning... | 2019-11-19 | Code |
| 5 | GDI-I3 | 17 | No | Generalized Data Distribution Iteration | 2022-06-07 | - |
| 6 | R2D2 | 15 | No | - | - | Code |
| 7 | LASER | 7 | No | Off-Policy Actor-Critic with Shared Experience R... | 2019-09-25 | - |
| 8 | Rainbow DQN | 4 | No | Rainbow: Combining Improvements in Deep Reinforc... | 2017-10-06 | Code |
| 9 | IMPALA, deep | 3 | No | IMPALA: Scalable Distributed Deep-RL with Import... | 2018-02-05 | Code |