| Mastering Atari, Go, Chess and Shogi by Planning with a Learned Model | ✓ Link | 631378.53 | | MuZero | 2019-11-19 |
| Agent57: Outperforming the Atari Human Benchmark | ✓ Link | 541280.88 | | Agent57 | 2020-03-30 |
| Online and Offline Reinforcement Learning by Planning with a Learned Model | ✓ Link | 374769.76 | | MuZero (Res2 Adam) | 2021-04-13 |
| Recurrent Experience Replay in Distributed Reinforcement Learning | ✓ Link | 315456.4 | | R2D2 | 2019-05-01 |
| Fully Parameterized Quantile Function for Distributional Reinforcement Learning | ✓ Link | 214060 | | Fearlessmrx | 2019-11-05 |
| Mastering Atari with Discrete World Models | ✓ Link | 11384 | | DreamerV2 | 2020-10-05 |
| Generalized Data Distribution Iteration | | 11330 | | GDI-H3(200M frames) | 2022-06-07 |
| Generalized Data Distribution Iteration | | 11330 | | GDI-H3 | 2022-06-07 |
| Generalized Data Distribution Iteration | | 10485 | | GDI-I3 | 2022-06-07 |
| GDI: Rethinking What Makes Reinforcement Learning Different From Supervised Learning | | 10485 | | GDI-I3 | 2021-06-11 |
| Distributed Prioritized Experience Replay | ✓ Link | 9328.6 | | Ape-X | 2018-03-02 |
| Train a Real-world Local Path Planner in One Hour via Partially Decoupled Reinforcement Learning and Vectorized Diversity | ✓ Link | 8616.4 | | ASL DDQN | 2023-05-07 |
| Dueling Network Architectures for Deep Reinforcement Learning | ✓ Link | 7413.0 | | Prior+Duel noop | 2015-11-20 |
| Self-Imitation Learning | ✓ Link | 6289.8 | | A2C + SIL | 2018-06-14 |
| #Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning | ✓ Link | 5214.0 | | TRPO-hash | 2016-11-15 |
| Dueling Network Architectures for Deep Reinforcement Learning | ✓ Link | 4672.8 | | Duel noop | 2015-11-20 |
| Distributional Reinforcement Learning with Quantile Regression | ✓ Link | 4384 | | QR-DQN-1 | 2017-10-27 |
| Prioritized Experience Replay | ✓ Link | 4380.1 | | Prior noop | 2015-11-18 |
| Implicit Quantile Networks for Distributional Reinforcement Learning | ✓ Link | 4324 | | IQN | 2018-06-14 |
| Deep Reinforcement Learning with Double Q-learning | ✓ Link | 4038.4 | | Prior+Duel hs | 2015-09-22 |
| A Distributional Perspective on Reinforcement Learning | ✓ Link | 3965.0 | | C51 noop | 2017-07-21 |
| Value Prediction Network | ✓ Link | 3811 | | VPN | 2017-07-11 |
| Prioritized Experience Replay | ✓ Link | 3510.0 | | Prior hs | 2015-11-18 |
| Learning values across many orders of magnitude | | 3469.6 | | DDQN+Pop-Art noop | 2016-02-24 |
| Increasing the Action Gap: New Operators for Reinforcement Learning | ✓ Link | 3248.96 | | Persistent AL | 2015-12-15 |
| Noisy Networks for Exploration | ✓ Link | 2923 | | NoisyNet-Dueling | 2017-06-30 |
| Count-Based Exploration in Feature Space for Reinforcement Learning | ✓ Link | 2770.1 | | Sarsa-φ-EB | 2017-06-25 |
| Model-Free Episodic Control with State Aggregation | | 2394 | 4020 | MFEC | 2020-08-21 |
| Dueling Network Architectures for Deep Reinforcement Learning | ✓ Link | 2332.4 | | Duel hs | 2015-11-20 |
| Increasing the Action Gap: New Operators for Reinforcement Learning | ✓ Link | 2305.82 | | Advantage Learning | 2015-12-15 |
| Deep Exploration via Bootstrapped DQN | ✓ Link | 2181.4 | | Bootstrapped DQN | 2016-02-15 |
| Dueling Network Architectures for Deep Reinforcement Learning | ✓ Link | 1683.3 | | DDQN (tuned) noop | 2015-11-20 |
| Deep Reinforcement Learning with Double Q-learning | ✓ Link | 1448.1 | | DDQN (tuned) hs | 2015-09-22 |
| Count-Based Exploration in Feature Space for Reinforcement Learning | ✓ Link | 1394.3 | | Sarsa-ε | 2017-06-25 |
| CURL: Contrastive Unsupervised Representations for Reinforcement Learning | ✓ Link | 924 | | CURL | 2020-04-08 |
| Deep Reinforcement Learning with Double Q-learning | ✓ Link | 797.4 | | DQN noop | 2015-09-22 |
| Evolving simple programs for playing Atari games | ✓ Link | 782 | | CGP | 2018-06-14 |
| Incentivizing Exploration In Reinforcement Learning With Deep Predictive Models | ✓ Link | 507.0 | | MP-EB | 2015-07-03 |
| Deep Reinforcement Learning with Double Q-learning | ✓ Link | 496.1 | | DQN hs | 2015-09-22 |
| Massively Parallel Methods for Deep Reinforcement Learning | ✓ Link | 426.6 | | Gorila | 2015-07-15 |
| Evolution Strategies as a Scalable Alternative to Reinforcement Learning | ✓ Link | 370.0 | | ES FF (1 hour) noop | 2017-03-10 |
| Human level control through deep reinforcement learning | ✓ Link | 328.3 | | Nature DQN | 2015-02-25 |
| DNA: Proximal Policy Optimization with a Dual Network Architecture | ✓ Link | 320 | | DNA | 2022-06-20 |
| IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures | ✓ Link | 317.75 | | IMPALA (deep) | 2018-02-05 |
| Policy Optimization With Penalized Point Probability Distance: An Alternative To Proximal Policy Optimization | ✓ Link | 316.87 | | POP3D | 2018-07-02 |
| Playing Atari with Six Neurons | ✓ Link | 300 | | IDVQ + DRSC + XNES | 2018-06-04 |
| The Arcade Learning Environment: An Evaluation Platform for General Agents | ✓ Link | 270.5 | | UCT | 2012-07-19 |
| The Arcade Learning Environment: An Evaluation Platform for General Agents | ✓ Link | 216.9 | | Best Learner | 2012-07-19 |
| Asynchronous Methods for Deep Reinforcement Learning | ✓ Link | 197.6 | | A3C LSTM hs | 2016-02-04 |
| Asynchronous Methods for Deep Reinforcement Learning | ✓ Link | 190.5 | | A3C FF hs | 2016-02-04 |
| []() | | 180.9 | | SARSA | |
| Asynchronous Methods for Deep Reinforcement Learning | ✓ Link | 180.1 | | A3C FF (1 day) hs | 2016-02-04 |
| Soft Actor-Critic for Discrete Action Settings | ✓ Link | 59.4 | | SAC | 2019-10-16 |