A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play
Open accessThe DOI matches its Crossref record; 30 of the 30 references it cites were found in a registry or an index; 5 of the 16 sources its claims cite were checked for support.
Abstract
One program to rule them all: Computers can beat humans at increasingly complex games, including chess and Go. However, these programs are typically constructed for a particular game, exploiting its properties, such as the symmetries of the board on which it is played. Silver et al. developed a program called AlphaZero, which taught itself to play Go, chess, and shogi (a Japanese version of chess) (see the Editorial, and the Perspective by Campbell). AlphaZero managed to beat state-of-the-art programs specializing in these three games. The ability of AlphaZero to adapt to various game rules is a notable step toward achieving a general game-playing system. Science , this issue p. 1140 ; see also pp. 1087 and 1118
Overview
CoChat Summary
TL;DR
AlphaZero is a single reinforcement-learning and MCTS algorithm that, given only game rules, learned superhuman play in chess, shogi, and Go through self-play. It defeated world-champion programs Stockfish, Elmo, and AlphaGo Zero using the same algorithm and network architecture across all three games.
Key findings
- In chess, AlphaZero defeated Stockfish 8, winning 155 games and losing 6 out of 1,000; it also won matches from common human openings and 2016 TCEC opening positions.
- In shogi, AlphaZero defeated Elmo, winning 98.2% of games as black and 91.2% overall; in Go, it defeated AlphaGo Zero, winning 61% of games.
- AlphaZero first outperformed Stockfish after 4 hours (300,000 steps), Elmo after 2 hours (110,000 steps), and AlphaGo Lee after 30 hours (74,000 steps); training ran for 700,000 steps, taking about 9 hours in chess, 12 hours in shogi, and 13 days in Go.
- AlphaZero searched only about 60,000 positions per second in chess and shogi, versus 60 million for Stockfish and 25 million for Elmo, yet still defeated Stockfish with 1/10 thinking time and won 46% against Elmo with 1/100 thinking time.
Methods
- AlphaZero uses a deep neural network \(f_\theta(s)\) that outputs move probabilities \(p\) and a scalar value \(v\), trained by reinforcement learning from self-play starting from random parameters; MCTS uses these outputs to guide search, with 800 simulations per move during training.
- The same algorithm, network architecture, and hyperparameters were used for chess, shogi, and Go, with no domain knowledge except game rules; no opening book, endgame tablebases, or domain-specific heuristics were used.
- Training used 5,000 first-generation TPUs to generate self-play games and 16 second-generation TPUs to train networks; evaluation matches used 3 hours per game plus 15 seconds per move, with opponents on 44 CPU cores and AlphaZero on four first-generation TPUs and 44 CPU cores.
Limitations
- Stated: AlphaZero did not use opening books, endgame tablebases, or domain-specific heuristics; the authors note that some such techniques could further improve performance but were left for future research.
- Stated: hyperparameters were reused across games without game-specific tuning, except for exploration noise and the learning rate schedule; matches against an opening book were low in diversity, and forced randomization increased losses.
- Evident from the text: results cover only three games with separate trained instances, no cross-game transfer is demonstrated, and evaluation compares AlphaZero on TPU hardware against CPU-based opponents, so hardware and time-control differences qualify direct strength comparisons.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
3 Valid · 2 Not valid · 8 Not checked · 43 Own results and interpretations
Every source the paper cites for this claim was checked and supports it.
In the paper
“Apart from the representation of positions and actions described above, AlphaZero uses the same network architecture as AlphaGo Zero ( 9 ), briefly recapitulated here.”
What the sources say
- Supportedcites [9] source
The Neural Network Architecture Comparison (p. 29) identifies AlphaGo Zero’s architecture as a residual tower with policy and value heads. The citation supplies the AlphaGo Zero architecture that the citing authors say AlphaZero uses; the comparison to AlphaZero is their own statement. (LLM, conf=0.88)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“and in Go, AlphaZero first outperformed AlphaGo Lee ( 9 ) after 30 hours (74,000 steps).”
What the sources say
- Partly supportedcites [9] sourceExperimental — accuracy not yet measured
Section 2, p. 7 reports that AlphaGo Zero outperformed AlphaGo Lee, but says this happened after 36 hours, not 30. The passage does not give 74,000 steps, and it names the system AlphaGo Zero rather than AlphaZero. (LLM, conf=0.90)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“The training algorithm achieved similar performance in all independent runs (see fig. S3), suggesting that the high performance of AlphaZero's training algorithm is repeatable.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“AlphaZero also defeated Stockfish when given 1 / 10 as much thinking time as its opponent (i.e. searching ~ 1 / 10 , 000 as many positions),”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“and won 46% of games against Elmo when given 1 / 100 as much time (i.e. searching ~ 1 / 40 , 000 as many positions), see Fig. 2.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
23 more claims
Create a free account to see all 56 claims and to verify the 3 references still open.
- Probably not a substantial claim. 10
- It may state two separate claims as one. 23
- The quote only partly states this claim. 1
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Games have historically served as highly effective metrics for evaluating progress in deep reinforcement learning research”
From BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games (2024) · cites this paper as Silver et al., 2018
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“In 2017, AlphaGo, using the same learning algorithms in TD-Gammon and a billion times more computing power, soundly beat Ke Jie, the world Go champion, in a game that is to chess what chess is to backgammon in complexity (Silver et al., 2018).”
From Large Language Models and the Reverse Turing Test (2023) · cites this paper as Silver et al., 2018
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“Most of the attention was on who won, but more important was how it won: AlphaGo was creative and had discovered new moves and board positions that were outside human experience.”
From Large Language Models and the Reverse Turing Test (2023) · cites this paper as Silver et al., 2018
Experimental — accuracy not yet measuredThe source backs part of the claim, not all of it — or backs it for a different population, condition or comparison.
In the citing paper
“The primary method for applying more computation at inference, common in the AlphaZero family of agents2,72, is scaling the tree search and running multiple independent search attempts.”
From Olympiad-level formal mathematical reasoning with reinforcement learning (2026) · cites this paper as [2]
Experimental — accuracy not yet measuredThe cited source says what the claim attributes to it.
In the citing paper
“Modern AI can write computer programs (Li et al., 2022) or play chess at super-human level (Silver et al., 2018), but the ability of AI to perceive and act in the world remains far below human level.”
From Scaling Instructable Agents Across Many Simulated Worlds (2024) · cites this paper as Silver et al., 2018
1 more claim about this paper
Create a free account to see every claim other papers make about this one.
Citations
Cites
Mastering the game of Go with deep neural networks and tree search
Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., van den Driessche, G., Schrittwieser, J., Antonoglou, I., Panneershelvam, V., Lanctot, M., et al. (2016) · Nature · doi:10.1038/nature16961 · Cited by 15,612
Self-citationAdvances in neural information processing systems 7
(1996) · Computers & Mathematics with Applications · doi:10.1016/s0898-1221(96)90463-0 · Cited by 14,281
Identity mappings in deep residual networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46493-0_38 · Cited by 10,741
Identity Mappings in Deep Residual Networks
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · arXiv · doi:10.48550/arxiv.1603.05027 · Cited by 10,665
Mastering the game of Go without human knowledge
Silver, D., Schrittwieser, J., Simonyan, K., Antonoglou, I., Huang, A., Guez, A., Hubert, T., Baker, L., Lai, M., Bolton, A., et al. (2017) · Nature · doi:10.1038/nature24270 · Cited by 9,084
Self-citationIn-Datacenter Performance Analysis of a Tensor Processing Unit
Jouppi, N. P., Young, C., Patil, N., Patterson, D., Agrawal, G., Bajwa, R., Bates, S., Bhatia, S., Boden, N., Borchers, A., et al. (2017) · Proceedings of the 44th Annual International Symposium on Computer Architecture · doi:10.1145/3079856.3080246 · Cited by 4,537
Encyclopedia of Artificial Intelligence
(2009) · IGI Global eBooks · doi:10.4018/978-1-59904-849-9 · Cited by 1,396
Campbell, M., Hoane, A. J., & Hsu, F.-H. (2002) · Artificial Intelligence · doi:10.1016/s0004-3702(01)00129-1 · Cited by 1,226
Lim, T.-S., Loh, W.-Y., & Shih, Y.-S. (2000) · Machine Learning · doi:10.1023/a:1007608224229 · Cited by 1,133
An analysis of alpha-beta pruning
Knuth, D. E., & Moore, R. W. (1975) · Artificial Intelligence · doi:10.1016/0004-3702(75)90019-3 · Cited by 1,004
Research works printed but not linked: 7
- C. J. Maddison A. Huang I. Sutskever D. Silver paper presented at the International Conference on Learning Representations 2015 San Diego CA 7 to 9 May 2015.
- D. N. L. Levy M. Newborn How Computers Play Chess (Ishi Press 2009).
- W. Steinitz The Modern Chess Instructor (Edition Olms 1990).
- O. Arenz “Monte Carlo chess ” master’s thesis Technische Universität Darmstadt (2012).
- 10.1023/a:1007634325138 · DOI
- T. Kaneko K. Hoki in Advances in Computer Games – 13th International Conference ACG 2011 Revised Selected Papers Tilburg Netherlands 20 to 22 November 2011 (Springer 2012) pp. 158–169.
- R. Ramanujan A. Sabharwal B. Selman in Proceedings of the 26th Conference on Uncertainty in Artificial Intelligence (UAI 2010) Catalina Island CA 8 to 11 July (AUAI Press 2010).
Cited by
Items 1,401–1,500 of 3,633
Reinforcement Learning for PID Gain Tuning in Ball-and-Beam Systems: A Comparative Study
Blake, X. J., & Aphiratsakun, N. (2024) · International Conference on Robotics, Engineering, Science, and Technology (RESTCON) · doi:10.1109/restcon60981.2024.10463594 · Cited by 9
“Applications for RL have been developed across diverse domains, such as mastering board games like Chess, Shogi, and Go [4], recommendation systems [5], energy management [6], healthcare [7], and computer vision [8].”
MethodEvaluating Superhuman Models with Consistency Checks
Fluri, L., Paleka, D., & Tramèr, F. (2024) · IEEE Conference on Secure and Trustworthy Machine Learning (SaTML) · doi:10.1109/satml59370.2024.00017 · Cited by 67
“To illustrate the challenge, consider a model trained to play chess—a canonical setting where models surpass humans [67, 17].”
BackgroundMethodInfluentialSarkhi, S. M. K., & Koyuncu, H. (2024) · AI · doi:10.3390/ai5030057 · Cited by 12
Learning quantum phases via single-qubit disentanglement
An, Z., Cao, C., Xu, C.-Q., & Zhou, D. L. (2024) · Quantum · doi:10.22331/q-2024-07-22-1421 · Cited by 12
What does AI’s success playing complex board games tell brain scientists?
Purves, D. (2019) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1909565116 · Cited by 8
Vision-Based Autonomous Driving: A Model Learning Approach
Baheri, A., Kolmanovsky, I. V., Girard, A., Tseng, H. E., & Filev, D. (2020) · Proceedings of the ... American Control Conference/Proceedings of the American Control Conference · doi:10.23919/acc45564.2020.9147510 · Cited by 12
“Reinforcement learning has shown dramatic success at learning to play Atari games and acting in complex environments to gain human-level control [9]–[11].”
BackgroundRoad Detection for Reinforcement Learning Based Autonomous Car
Holen, M., Saha, R., Goodwin, M., Omlin, C., & Sandsmark, K. E. (2020) · International Conferences on Information Science and System · doi:10.1145/3388176.3388199 · Cited by 47
The Genius of the 'Original Imitation Game' Test
Sterrett, S. G. (2020) · Minds and Machines · doi:10.1007/s11023-020-09543-6 · Cited by 8
“(Silver et al. 2018).”
Shtanko, A., Kulik, S. D., & Kondakov, A. (2020) · Journal of Physics Conference Series · doi:10.1088/1742-6596/1685/1/012011 · Cited by 8
Towards a common environment for learning scheduling algorithms
Cunha, R. L. F., & Chaimowicz, L. (2020) · Proceedings - International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunication Systems · doi:10.1109/mascots50786.2020.9285940 · Cited by 8
“Particularly for games, Reinforcement Learning (RL) has had compelling, super-human success stories both in board games, such as Backgammon [26], Chess [24], and Go [24], and in digital ones, such as in games from the Atari system [1, 9].”
BackgroundAlgorithmic and human prediction of success in human collaboration from visual features
Saveski, M., Awad, E., Rahwan, I., & Cebrián, M. (2021) · Scientific Reports · doi:10.1038/s41598-021-81145-3 · Cited by 12
Artificial and Human Intelligence in Mental Health
Sigman, M., Slezak, D. F., Drucaroff, L. J., Ribeiro, S. T. G., & Carrillo, F. (2021) · AI Magazine · doi:10.1002/j.2371-9621.2021.tb00009.x · Cited by 8
“More recently, however, as is widely known, the approach to Ai in chess changed dramatically, using remarkable computer power and a great insight of doing convolutions in arbitrary deep spaces, to promote a process of self-discovery (Campbell, Joseph Hoane, and Hsu 2002; Silver et al. 2018).”
BackgroundInfluentialEvaluating Critical Reinforcement Learning Framework in the Field
Ju, S., Zhou, G., Abdelshiheed, M., Barnes, T. M., & Chi, M. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-78292-4_18 · Cited by 13
Vector Quantization for Adaptive State Aggregation in Reinforcement Learning
Mavridis, C. N., & Baras, J. S. (2021) · Proceedings of the ... American Control Conference/Proceedings of the American Control Conference · doi:10.23919/acc50511.2021.9483052 · Cited by 8
“Reinforcement learning and value function approximation algorithms have been receiving increasing attention recently and have resulted in impressive applications in multiple fields including elevator control, robot soccer [1], and game-playing, such as Backgammon, Chess and Go [2], [3].”
BackgroundGao, Y., Wu, L., & Li, H. (2021) · IEEE Conference on Games (CoG) · doi:10.1109/cog52621.2021.9619111 · Cited by 8
“…learning, mainly based on reinforcement learning, has become a benchmarking method of many board games and been widely used in training other board game AI like ELF OpenGo [1], Polygames [2], and KataGo [3], since the birth of the AlphaGo family (AlphaGo [4], AlphaGo Zero [5], and Alpha Zero [6]).”
MethodOlivaw: Mastering Othello Without Human Knowledge, nor a Fortune
Norelli, A., & Panconesi, A. (2022) · IEEE Transactions on Games · doi:10.1109/tg.2022.3157345 · Cited by 8
“Although cheaper in some ways, the agents in [2]–[5] still use thousands of GPUs or hundreds of TPUs to master board games.”
BackgroundAIR5: Five Pillars of Artificial Intelligence ResearchPreprint
Ong, Y.-S., & Gupta, A. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1812.11509 · Cited by 8
“2928344 of chess [6], and Carnegie Mellon University’s AI defeating four of the”
BackgroundDynamic Origin-Destination Matrix Prediction with Line Graph Neural Networks and Kalman FilterPreprint
Xiong, X., Özbay, K. M. A., Li, J., et al. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1905.00406 · Cited by 8
“In recent years, deep neural networks have shown to be effective in approximating nonlinear features in classification, regression and control problems [18, 19].”
Efficient Learning of Distributed Linear-Quadratic ControllersPreprint
Fattahi, S., Matni, N., & Sojoudi, S. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1909.09895 · Cited by 8
“…of machine learning (ML) techniques applied to complex decision making problems [1] such as image classification [2], video and board games [3, 4, 5], and robotics [6, 7, 8], the use of ML for the control of autonomous systems interacting with physical environments has been an active area of…”
BackgroundRobot Navigation with Map-Based Deep Reinforcement LearningPreprint
Chen, G., Pan, L., Chen, Y., et al. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2002.04349 · Cited by 8
“On the other hands, deep reinforcement learning (DRL) methods have achieved remarkable success in many challenging tasks [9]–[11].”
BackgroundReinforcement Learning Generalization with Surprise MinimizationPreprint
Chen, J. Z. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2004.12399 · Cited by 22
“Most notably, it has beaten human players in board games [1, 2] and reached expert level in video games [3, 4].”
BackgroundScalable Anytime Planning for Multi-Agent MDPsPreprint
Choudhury, S., Gupta, J. K., Morales, P., et al. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2101.04788 · Cited by 8
“Similarly, extending ideas from Alpha-Zero [2, 37] would be particularly relevant for distilling the coordinated individual actions for the agents into decentralized policies [34] with FV-MCTS acting as a scalable policy improvement operator [16].”
BackgroundA deep learning theory for neural networks grounded in physicsPreprint
Scellier, B. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2103.09985 · Cited by 34
AndroidEnv: A Reinforcement Learning Platform for AndroidPreprint
Toyama, D., Hamel, P., Gergely, A., Comanici, G., Glaese, A., Ahmed, Z., Jackson, T., Mourad, S., & Precup, D. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2105.13231 · Cited by 126
“RL methods have demonstrated success not only in game playing, for example checkers (Schaeffer et al., 1992), chess (Campbell et al., 2002; Silver et al., 2018), Go (Silver et al., 2016), poker (Moravčík et al., 2017), Atari (Mnih et al., 2015) and Starcraft II (Vinyals et al., 2019), but also in…”
MethodReward Shaping to Improve the Performance of Deep Reinforcement Learning in Inventory ManagementPreprint
Moor, B. J. D., Gijsbrechts, J., & Boute, R. N. (2021) · SSRN Electronic Journal · doi:10.2139/ssrn.3804655 · Cited by 6
“…Lee Sedol in the complex game of GO (Silver et al., 2016); the creation of AlphaZero, a better and more general-purpose algorithm than AlphaGo (Silver et al., 2018); and the development of MuZero, achieving superior performance in GO while also excelling in other games like chess by learning…”
Human-guided reinforcement learning: methodology and application to autonomous driving
Wu, J. (2023) · DR-NTU (Nanyang Technological University) · doi:10.32657/10356/169780 · Cited by 8
Algorithmic Collusion or Competition: the Role of Platforms' Recommender SystemsPreprint
Xu, X., Lee, S., & Tan, Y. (2023) · SSRN Electronic Journal · doi:10.2139/ssrn.4579458 · Cited by 12
Not so disruptive yet? Characteristics, distribution and determinants of robots in Europe
Fernández‐Macías, E., Klenert, D., & Antón, J. (2021) · Structural Change and Economic Dynamics · doi:10.1016/j.strueco.2021.03.010 · Cited by 102
Induction of Subgoal Automata for Reinforcement Learning
Furelos-Blanco, D., Law, M., Russo, A., Broda, K. B., & Jönsson, A. (2020) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v34i04.5802 · Cited by 7
“2015) or mastering complex board games (Silver et al. 2018).”
Efficient Course Recommendation using Deep Transformer based Ensembled Attention Model
Madhavi, A., Nagesh, A., & Govardhan, A. (2023) · ICST Transactions on e-Education and e-Learning · doi:10.4108/eetel.4470 · Cited by 9
“The use of RL has spread widely across several fields [36-39].”
Reinforcement Learning as an Approach to Train Multiplayer First-Person Shooter Game Agents
Almeida, P., Carvalho, V. H., & Simões, A. (2024) · Technologies · doi:10.3390/technologies12030034 · Cited by 10
Realizing the Carbon-Aware Service Provision in ICT System
Sun, P., Lan, J., Hu, Y., Guo, Z., Wu, C., & Wu, J. (2024) · IEEE Transactions on Network and Service Management · doi:10.1109/tnsm.2024.3385484 · Cited by 8
“For example, AlphaZero proposed by DeepMind [52] won top human Go players.”
MethodLi, M., Li, N., Shao, X., Wang, J., & Xu, D. (2024) · CAAI Artificial Intelligence Research · doi:10.26599/air.2024.9150033 · Cited by 12
Heads-Up Multitasker: Simulating Attention Switching On Optical Head-Mounted Displays
Bai, Y., Ikkala, A., Oulasvirta, A., Zhao, S., Wang, L. J., Yang, P., & Xu, P. (2024) · CHI Conference on Human Factors in Computing Systems · doi:10.1145/3613904.3642540 · Cited by 10
Enhanced LSTM‐DQN algorithm for a two‐player zero‐sum game in three‐dimensional space
Lu, B., Ru, L., Lv, M., Hu, S., Zhang, H., & Zhao, Z. (2024) · IET Control Theory and Applications · doi:10.1049/cth2.12677 · Cited by 8
Machine Learning Algorithm Guides Catalyst Choices for Magnesium‐Catalyzed Asymmetric Reactions
Baczewska, P., Kulczykowski, M., Zambroń, B. K., Jaszczewska‐Adamczak, J. A., Pakulski, Z., Roszak, R., Grzybowski, B. A., & Młynarski, J. (2024) · Angewandte Chemie · doi:10.1002/ange.202318487 · Cited by 7
Bridge Bidding via Deep Reinforcement Learning and Belief Monte Carlo Search
Qiu, Z., Wang, S., You, D., & Zhou, M. (2024) · IEEE/CAA Journal of Automatica Sinica · doi:10.1109/jas.2024.124488 · Cited by 8
Reinforcement learning for fuselage shape control during aircraft assembly
Lutz, T., Wang, Y., Yue, X., & Camelio, J. A. (2024) · IISE Transactions · doi:10.1080/24725854.2024.2413136 · Cited by 9
“The proposed method can be regarded as a trial for developing an “AlphaGo” or “AlphaZero” for advanced aircraft assembly, in particular, the fuselage shape control process.”
Optimizing Tensor Computation Graphs with Equality Saturation and Monte Carlo Tree Search
Hartmann, J., He, G., & Yoneki, E. (2024) · International Conference on Parallel Architectures and Compilation Techniques (PACT) · doi:10.1145/3656019.3689611 · Cited by 14
Yi, M., Wang, X., Liu, J., Zhang, Y., & Hou, R. (2025) · IEEE Transactions on Communications · doi:10.1109/tcomm.2025.3543185 · Cited by 10
“Over the past few years, the DRL method has received significant attention due to its impressive success in intricate tasks such as international Go, games, and controlling complex machinery for various operations [36], [37], [38].”
Power grid fault diagnosis based on an improved BERT model
Zheng, Y., Zheng, Y., Zhang, Z., Liu, B., Guo, Z., & Wei, J. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.128648 · Cited by 8
Picking strategies in games of cooperation
García, J., & Traulsen, A. (2025) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2319925121 · Cited by 14
Reinforcement learning in artificial intelligence and neurobiology
Alkam, T., Benschoten, A. H. V., & Tarshizi, E. (2025) · Neuroscience Informatics · doi:10.1016/j.neuri.2025.100220 · Cited by 12
Xue, J., Li, X., & Zhang, N. (2025) · Robotics and Computer-Integrated Manufacturing · doi:10.1016/j.rcim.2025.103159 · Cited by 9
“It is applied in various domains, such as gaming [63], competition [64], and large language models[55].”
Rearchitecting in-memory object stores for low latency
Zhuo, D., Zhang, K., Li, Z., Zhuang, S., Wang, S., Chen, A., & Stoica, I. G. (2021) · Proceedings of the VLDB Endowment · doi:10.14778/3494124.3494138 · Cited by 8
“Monte Carlo Tree Search (MCTS) [16] is a heuristic tree search algorithm, which is proved to be successful in gameplay [24, 28, 65, 66].”
BackgroundLuo, Q., Tan, T.-P., Su, Y., & Jin, Z. (2022) · IEEE Transactions on Games · doi:10.1109/tg.2022.3223926 · Cited by 8
“For example, AlphaGo [1], AlphaZero [2], and MuZero [3] have achieved state-of-the-art performance in Go [1], [2], [3], a strategy board game.”
BackgroundStorehouse: a Reinforcement Learning Environment for Optimizing Warehouse Management
Cestero, J., Quartulli, M., Metelli, A. M., & Restelli, M. (2022) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn55064.2022.9891985 · Cited by 15
A Reinforcement Learning Based Slope Limiter for Second‐Order Finite Volume Schemes
Schwarz, A., Keim, J., Chiocchetti, S., et al. (2023) · PAMM · doi:10.1002/pamm.202200207 · Cited by 7
Chromothripsis detection with multiple myeloma patients based on deep graph learning
Yu, J., Chen, N., Zheng, Z., Gao, M. Y., Liang, N., & Wong, K. (2023) · Bioinformatics · doi:10.1093/bioinformatics/btad422 · Cited by 12
“Deep learning (DL) algorithms powered by advanced computing units have outperformed humans in visual tasks, such as playing Atari games (Mnih et al. 2015) and strategic games, such as Go (Silver et al. 2018).”
Chess Game to Improve the Mental Ability of Alzheimer's Patients using A3C
Joypriyanka, M., & Surendran, R. (2023) · International Conference on Electrical, Computer and Communication Technologies (ICECCT) · doi:10.1109/icecct56650.2023.10179809 · Cited by 12
Go, S., Jiang, Y., & Loke, D. K. (2023) · Advanced Intelligent Systems · doi:10.1002/aisy.202300335 · Cited by 14
Leite, S., Mota, B., Silva, A. R., Commons, M. L., Miller, P. M., & Rodrigues, P. P. (2023) · PLoS ONE · doi:10.1371/journal.pone.0290743 · Cited by 8
“Currently, neural network models designed to solve complex problems are usually programmed with a complex structure built from scratch, which is usually employed for solving the complex problem from scratch [29–32] using enormous datasets [33], and tending to overfit [32].”
BackgroundTumasyan, A., Adam, W., Andrejkovic, J. W., et al. (2023) · Physical review. D/Physical review. D. · doi:10.1103/physrevd.108.052002 · Cited by 7
Pessimistic value iteration for multi-task data sharing in Offline Reinforcement Learning
Bai, C., Wang, L., Hao, J., Yang, Z., Zhao, B., Wang, Z., & Li, X. (2023) · Artificial Intelligence · doi:10.1016/j.artint.2023.104048 · Cited by 16
“DRL has achieved remarkable success in a variety of tasks, including Atari games [2], Go [3], and StarCraft [4].”
MethodA Survey: Navigating the Landscape of Incremental Learning Techniques and Trends
Li, Y., Wu, W., Luo, X., Zheng, M., Zhang, Y., & Peng, B. (2023) · International Conference on Intelligent Systems and Knowledge Engineering (ISKE) · doi:10.1109/iske60036.2023.10481497 · Cited by 7
“Traditional deep learning models, after offline training, can exhibit performance surpassing human-level in tasks such as Atari games [1] and visual tasks [2].”
MethodAutomated Computation of Therapies Using Failure Mode and Effects Analysis in the Medical Domain
Luttermann, M., Baake, E., Bouchagiar, J., Gebel, B., Grüning, P., Manikwadura, D., Schollemann, F., Teifke, E., Rostalski, P., & Möller, R. (2024) · Künstliche Intell. · doi:10.1007/s13218-023-00810-z · Cited by 8
“Similar to the AlphaZero program [24] (which has been developed to master the games of chess, shogi, and go where the state spaces are also huge), one could use reinforcement learning to learn an approximation of the optimal policy.”
Songyang, L. (2024) · Measurement Sensors · doi:10.1016/j.measen.2024.101208 · Cited by 7
Learning in games: a systematic review
Qin, R.-J., & Yu, Y. (2024) · Science China Information Sciences · doi:10.1007/s11432-023-3955-x · Cited by 12
“Following studies AlphaGoZero [196] and AlphaZero [197] conquer Go and other board games without human initialization.”
MethodFrom predicting to decision making: Reinforcement learning in biomedicine
Liu, X., Zhang, J., Hou, Z., Yang, Y., & Gao, Y. Q. (2024) · Wiley Interdisciplinary Reviews Computational Molecular Science · doi:10.1002/wcms.1723 · Cited by 9
Examining the applicability of the Protection of Personal Information Act in AI-driven environments
Mbonye, V., Moodley, M., & Nyika, F. D. (2024) · South African journal of information management · doi:10.4102/sajim.v26i1.1808 · Cited by 20
“AlphaZero won 28 games, lost none, and tied for 72 others against Stockfish in a timed 100-game tournament (Silver et al. 2018).”
Reinforcement learning with predefined and inferred reward machines in stochastic games
Hu, J., Paliwal, Y., Kim, H., Wang, Y., & Xu, Z. (2024) · Neurocomputing · doi:10.1016/j.neucom.2024.128170 · Cited by 10
Z-Score Experience Replay in Off-Policy Deep Reinforcement Learning
Yang, Y., Xi, M., Dai, H., Wen, J., & Yang, J. (2024) · Sensors · doi:10.3390/s24237746 · Cited by 7
CuAsmRL: Optimizing GPU SASS Schedules via Deep Reinforcement Learning
He, G., & Yoneki, E. (2025) · Proceedings of the CGO/Proceedings of the ... CGO · doi:10.1145/3696443.3708943 · Cited by 17
Practical Open-Loop Optimistic Planning
Leurent, E., & Maillard, O.-A. (2020) · Lecture notes in computer science · doi:10.1007/978-3-030-46133-1_5 · Cited by 9
“Despite its popularity [15,17,16], UCT has been shown to suffer from several limitations: its sample complexity can be at least doubly-exponential for some problems (e.g. when a narrow optimal path is hidden in a suboptimal branch), which is much worse than uniform planning [7].”
BackgroundThe $$\alpha \mu $$ Search Algorithm for the Game of Bridge
Cazenave, T., & Ventos, V. (2021) · Communications in computer and information science · doi:10.1007/978-3-030-89453-5_1 · Cited by 11
“As superhuman level has been reached for Go starting from zero knowledge [18] and as it is also the case for other two player complete information games such as Chess and Shogi [17] some of the next challenges in games are imperfect information games such as Bridge or Poker.”
Straight to the Point : Fast-forwarding Videos via Reinforcement Learning Using Textual Data
Ramos, W., Silva, M. M., Junior, E. R. A., et al. (2020) · Lancaster EPrints (Lancaster University) · Cited by 7
Trained Model Reuse of Autonomous-Driving in Pygame with Deep Reinforcement Learning
Guo, Y., Gao, Q., & Pan, F. (2020) · Chinese Control Conference (CCC) · doi:10.23919/ccc50068.2020.9188547 · Cited by 7
“…later, DeepMind company combines Deep Learning with Q-learning, which is Deep Q-Network (DQN), and is applied in Atari games [7] and AlphaGo Zero [8] autonomous driving, Sallab AE et al. demonstrated that reinforcement learning can be applied in autonomous driving [9], then Yu, A. et al.…”
MethodPathways to controlled 3D deformation of graphene: Manipulating the motion of topological defects
Annevelink, E. J., Johnson, H. T., & Ertekin, E. (2021) · Current Opinion in Solid State and Materials Science · doi:10.1016/j.cossms.2020.100893 · Cited by 7
Monte-Carlo Planning and Learning with Language Action Value Estimates
Jang, Y., Seo, S., Lee, J., & Kim, K.-E. (2021) · Cited by 8
Evolving Evaluation Functions for Collectible Card Game AI
Miernik, R., & Kowalski, J. (2022) · International Conference on Agents and Artificial Intelligence (ICAART) · doi:10.5220/0010806900003116 · Cited by 9
Assessing Explainability in Reinforcement Learning
Zelvelder, A. E., Westberg, M., & Främling, K. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-82017-6_14 · Cited by 13
Two-stage training algorithm for AI robot soccer
Kim, T., Vecchietti, L. F., Choi, K., Sarıel, S., & Har, D. (2021) · PeerJ Computer Science · doi:10.7717/peerj-cs.718 · Cited by 9
“Recently, deep reinforcement learning (DRL) has been widely applied to deterministic games [1], video games [2, 3, 4], and complex robotic tasks [5, 6, 7, 8].”
BackgroundAvoiding collaborative paradox in multi‐agent reinforcement learning
Kim, H., Kim, S., Lee, D., & Jang, I. (2021) · ETRI Journal · doi:10.4218/etrij.2021-0010 · Cited by 8
“…the great success of RL combined with deep learning on single-agent applications [19], such as video games [20] and traditional complex games [21], there remain new challenges for MARL in real-world applications, such as self-control drones [22], self-driving cars [23], and self-operating…”
BackgroundSpatial Action Maps Augmented with Visit Frequency Maps for Exploration Tasks
Wang, Z., & Papanikolopoulos, N. P. (2021) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros51168.2021.9636813 · Cited by 9
Multi-agent reinforcement learning for character control
Li, C., Fussell, L., & Komura, T. (2021) · The Visual Computer · doi:10.1007/s00371-021-02269-1 · Cited by 8
“Alp-haZero [38] use deep neural networks with Monte Carlo search trees to build an agent and trained it by self-play.”
BackgroundHe, K., Wu, H., Wang, Z., & Li, H. (2022) · International Journal of Intelligent Systems · doi:10.1002/int.22837 · Cited by 7
“In recent years, deep reinforcement learning has been applied to find the near ‐ optimal strategies in perfect information games such as chess, go, shogi, 1,2 and so on.”
Skills to Drive: Successor Features for Autonomous Highway Pilot
Szőke, L., Aradi, S., Bécsi, T., & Gáspár, P. (2022) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2022.3150493 · Cited by 7
“Since then, other successful implementations and agents have been published in even more complex domains, see [1]–[5].”
BackgroundScalable Online Planning for Multi-Agent MDPs
Choudhury, S., Gupta, J. K., Morales, P., & Kochenderfer, M. J. (2022) · Journal of Artificial Intelligence Research · doi:10.1613/jair.1.13261 · Cited by 7
Jaskie, K., & Spanias, A. (2022) · Synthesis lectures on artificial intelligence and machine learning · doi:10.2200/s01152ed1v01y202111aim051 · Cited by 20
Attitude Control for Fixed-Wing Aircraft Using Q-Learning
Richter, D. J., Natonski, L. J., Shen, X., & Calix, R. A. (2022) · Lecture notes in computer science · doi:10.1007/978-3-030-98404-5_59 · Cited by 8
“From So Simple a Beginning”: Species of Artificial Intelligence
Shadbolt, N. (2022) · Daedalus · doi:10.1162/daed_a_01898 · Cited by 10
MoGym: Using Formal Models for Training and Verifying Decision-making Agents
Gros, T. P., Hermanns, H., Hoffmann, J., Klauck, M., Köhl, M. A., & Wolf, V. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-13188-2_21 · Cited by 14
“Given the successes of deep RL across many diverse environments [58,57,54,23,49,1,47], one is tempted to expect it to work well on the considered environments [31,22], too.”
BackgroundA Survey of Reinforcement Learning Toolkits for Gaming: Applications, Challenges and Trends
Jayaramireddy, C. S., Naraharisetti, S. V. V. S. S., Nassar, M., & Mekni, M. (2022) · Lecture notes in networks and systems · doi:10.1007/978-3-031-18461-1_11 · Cited by 11
A novel control strategy of automatic parallel parking system based on Q-learning
Zhang, L., Wu, J., Liu, C., et al. (2022) · Proceedings of the Institution of Mechanical Engineers Part D Journal of Automobile Engineering · doi:10.1177/09544070221137406 · Cited by 7
“24 The complex nonlinear problems often appear in engineering, whose mathematical model is often difficult to establish.”
BackgroundModeling opponent learning in multiagent repeated games
Hu, Y., Han, C., Li, H., et al. (2022) · Applied Intelligence · doi:10.1007/s10489-022-04249-x · Cited by 7
“A famous study concerned the AlphaZero algorithm [18], which beat the top human players in Go.”
BackgroundFricke, C., Wolff, D., Kemmerling, M., & Elgeti, S. (2023) · Advances in Computational Science and Engineering · doi:10.3934/acse.2023001 · Cited by 7
“At the latest after its breakthrough in playing Atari games [21], RL has become an established method for solving tasks where a valid solution strategy can be learned from repeated interactions, e.g., like playing computer games [40] or tabletop games [36].”
MethodA case study for unlocking the potential of deep learning in asset-liability-management
Krabichler, T., & Teichmann, J. (2023) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2023.1177702 · Cited by 11
“Particularly, the empirical approach with static replicating portfolios described in the Section 4.2 of Spillmann et al. (2019) is often applied in mid-sized retail banks.”
BackgroundMethodInfluentialWhy Deep Learning’s Performance Data Are Misleading
Weng, J. (2023) · International Conference on Artificial Intelligence in Electronics Engineering (AIEE) · doi:10.1145/3586185.3586188 · Cited by 13
Homing Guidance Law Design against Maneuvering Targets Based on DDPG
Liang, Y., Tang, J., Bai, Z., & Li, K. (2023) · International Journal of Aerospace Engineering · doi:10.1155/2023/4188037 · Cited by 7
“The DRL method can effectively overcome the difficulties of complex space and high dimensions [19, 20], so it may have advantages in homing guidance.”
MethodOffline Quantum Reinforcement Learning in a Conservative Manner
Cheng, Z., Zhang, K., Shen, L., & Tao, D. (2023) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v37i6.25872 · Cited by 11
“Reinforcement Learning (RL) RL, one of the three domains of machine learning (ML), has attracted extensive interest due to its ability to solve complicated decision-making tasks with superhuman performance (Silver et al. 2018).”
BackgroundReinforcement Learning Applied to AI Bots in First-Person Shooters: A Systematic Review
Almeida, P., Carvalho, V. H., & Simões, A. (2023) · Algorithms · doi:10.3390/a16070323 · Cited by 9
Real World Offline Reinforcement Learning with Realistic Data Source
Zhou, G., Ke, L., Srinivasa, S. S., Gupta, A., Rajeswaran, A., & Kumar, V. (2023) · Proceedings - IEEE International Conference on Robotics and Automation/Proceedings · doi:10.1109/icra48891.2023.10161474 · Cited by 27
“Despite rapid advances, the applicability of Deep Reinforcement Learning (DRL) algorithms [1, 2, 3, 4, 5, 6, 7, 8] to real-world robotics tasks is limited due to sample inefficiency and safety considerations.”
BackgroundMixture of personality improved spiking actor network for efficient multi-agent cooperation
Li, X., Ni, Z., Ruan, J., Meng, L., Shi, J., Zhang, T., & Xu, B. (2023) · Frontiers in Neuroscience · doi:10.3389/fnins.2023.1219405 · Cited by 9
“Although traditional self-play methods (Silver et al., 2018) have achieved significant advantages and can often converge to an optimal equilibrium strategy in competitive games, they tend to overfit specific partners for cooperative tasks.”
BackgroundMethodTing, D. S. W., Humayun, M. S., & Huang, S. S. (2023) · Current Opinion in Ophthalmology · doi:10.1097/icu.0000000000000984 · Cited by 10
Hicks, C., Mavroudis, V., Foley, M., Davies, T., Highnam, K., & Watson, T. (2023) · ACM Workshop on Artificial Intelligence and Security (AISec) · doi:10.1145/3605764.3623986 · Cited by 23
Plante, M. (2023) · Frontiers in Bioengineering and Biotechnology · doi:10.3389/fbioe.2023.1266298 · Cited by 9
“For example, computer programs that would enable the design of virtual cells (Tomita et al., 2009; Schellenberger et al., 2011; O ’ Brien et al., 2015; Breuer et al., 2019) or arti fi cial intelligence supporting living properties (e.g.,: AlphaGo developed by Silver et al., 2018).”
BackgroundMethodLi, W., Huang, J., Su, Q., et al. (2023) · Computer Communications · doi:10.1016/j.comcom.2023.11.027 · Cited by 7
DanZero: Mastering GuanDan Game with Reinforcement Learning
Lu, Y., Zhao, J. Q., Zhao, Y., Zhou, W., & Li, H. (2023) · IEEE Conference on Games (CoG) · doi:10.1109/cog57401.2023.10333179 · Cited by 9
BCQQ: Batch-Constraint Quantum Q-Learning with Cyclic Data Re-uploading
Periyasamy, M., Hölle, M., Wiedmann, M., Scherer, D. D., Plinge, A., & Mutschler, C. (2024) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn60899.2024.10651268 · Cited by 15
“Using this framework, agents based on deep neural networks (DNNs) have been remarkably successful in a variety of complex tasks, including super-human performance in computationally-hard board games [1] and discovering fast matrix multiplication algorithms [2].”
MethodReCoRe: Regularized Contrastive Representation Learning of World Model
Poudel, R. P. K., Pandya, H., Liwicki, S., & Cipolla, R. (2024) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr52733.2024.02161 · Cited by 21
“In recent years, deep RL algorithms have been successfully employed for designing optimal strategies for games [25, 31] and shown a promise for controlling robots [21, 35].”
BackgroundMethod
Metadata
- Authors
- Silver, D., Hubert, T., Schrittwieser, J., Antonoglou, I., Lai, M., Guez, A., Lanctot, M., Sifre, L., Kumaran, D., Graepel, T., Lillicrap, T., Simonyan, K., & Hassabis, D.
- Affiliations
- Google DeepMind (United Kingdom) (GB)
- Corresponding author
- Silver, David
- Hubert, Thomas
- Schrittwieser, Julian
- Hassabis, Demis
- ORCID
- Venue
- Science
- Published
- 2018-12-07
- Volume
- 362
- Issue
- 6419
- Pages
- 1140-1144
- ISSN
- 0036-8075 (ISSN-L)
- 1095-9203
- OpenAlex
- W2902907165
- Cited by
- 3,562
- References
- 30
- Citations by year
- 2026: 257 · 2025: 427 · 2024: 512 · 2023: 514 · 2022: 569 · 2021: 546 · 2020: 462 · 2019: 256 · 2018: 13
- FWCI
- 179.89
- Type
- journal article
- Language
- en
- Open access
- bronze
- doi.org
- Locations
- Science (publishedVersion)
- PubMed (publishedVersion)
- UCL Discovery (University College London) (submittedVersion)
- Funders
- DeepMind
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence in Games
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Artificial Intelligence › Artificial Intelligence Applications
- Keywords
- AlphaZero, chess, shogi, Go, evaluation function learning, game tree search
- MeSH terms
- Algorithms, Artificial Intelligence, Humans, Reinforcement, Psychology, Software, Video Games
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-10 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.