The 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.
A general reinforcement learning algorithm that masters chess, shogi, and Go through self-play
Open accessAbstract
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 3,301–3,400 of 3,622
Gamifying the Vehicle Routing Problem with Stochastic Requests
Kullman, N., Dudorov, N., Cousineau, M., Mendoza, J. E., & Goodson, J. C. (2025) · INFORMS journal on computing · doi:10.1287/ijoc.2024.0838 · Cited by 4
“Perhaps the most well-known application is to games, where DRL has achieved superhuman performance in chess (Silver et al., 2018), Go (Silver et al., 2018), Doom (Lample and Chaplot, 2017), Texas Hold’em Poker (Heinrich and Silver, 2016), and StarCraft II (Vinyals et al., 2019).”
Wang, H., Zhang, X., & Mu, C. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5611353
Kim, M., Park, J.-C., Choi, S., & Kim, G.-W. (2025) · Information · doi:10.3390/info16100898
“…degrade planning quality, potentially causing fatal decision errors during high-speed driving; and (iii) reflecting complex vehicle dynamics, sensor noise, and road friction variations requires massive parameters and computations, making it difficult to meet real-time control cycles [10,11].”
Grundlagen der Künstlichen Intelligenz
Plafky, C. S., & Badertscher, H. (2025) · Springer eBooks · doi:10.1007/978-3-658-49009-6_3
A Dharmakīrtian Model of Relevance Realization in Cognitive Agents
Amir, N., & Dunne, J. D. (2025) · Contemporary Buddhism · doi:10.1080/14639947.2025.2564476 · Cited by 2
“…the cumulative reward signal expected over repeated interactions between the agent and its environment have been successfully used to design artificial systems that master games such as Chess or Go (Silver et al. 2018), as well as a variety of other cognitively demanding tasks (Silver et al. 2021).”
Learning-Based Recommendations for Efficient Urban Visual Query
Wu, Z., Chen, W., Wu, X., Zhou, Z., Feng, Y., Lu, J., Zhou, Z., & Xu, M. (2025) · IEEE Transactions on Visualization and Computer Graphics · doi:10.1109/tvcg.2025.3625071
“The agent learns recommending policies following the approach employed in AlphaGo Zero [10]: constructing the action space of RL with feasible query subspace; forming the state space of RL with the current query results; and regarding the significance measure of query results as the reward of the RL.”
InfluentialHigh-Density Chess Encryption: A CNN and DQN Framework for SteganographyPreprint
Rahman, M. D. M., Das, S., & Shamir, A. (2025) · Research Square · doi:10.21203/rs.3.rs-7783101/v1
Testing Obedience and Control in AGI: Exploring Irrational Commands and the AI Control Problem
Yampolskiy, R. V., & Field, S. (2025) · Springer eBooks · doi:10.1007/978-981-95-3658-0_69
Creating Adaptive Adversaries for Fighting Games Using Deep Reinforcement Learning
Kamran, U., & Mujtaba, H. (2025) · International Conference on Communication, Computing and Digital Systems (C-CODE) · doi:10.1109/c-code67372.2025.11204120
“Actor-Critic methods rely on two components: the Critic, a value function that estimates the quality of an action given a state, and the Actor, which selects actions based on expected returns and updates its policy using feedback from the Critic [4].”
Do Capacity Constrained Bots Collude?∗Preprint
Lilletvedt, J. S., & Dyskeland, O. K. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5690543
Semantic Modeling of Behavior in Users of Online Communities
Komardin, M. A., Panilov, P. A., & Tsibizova, T. Y. (2025) · doi:10.1109/mlsd65526.2025.11220709
“It can be used to identify key actors in network interaction, track the dynamics of public opinion, and predict user behavior [2].”
Reinforcement learning for accelerator beamline control: A simulation-based approach
Ibrahim, A., Petrenko, A., Kaledin, M., et al. (2026) · International Journal of Modern Physics E · doi:10.1142/s0218301326410272
Learning Action Strategies in the Wumpus World with DQN
Draszawka, K., Szymański, J., Gil, D., Pont, M. T. S., & Mora, H. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-09321-9_28
Optimizing printing processes with MCTS
Kukk, K., Torim, A., Eessaar, E., et al. (2025) · Intelligent Systems with Applications · doi:10.1016/j.iswa.2025.200602
Han, Y., Zhang, L., Meng, D., Zhang, Z., Hu, X., & Weng, S. (2025) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2025.3626399 · Cited by 2
“Monte Carlo Tree Search (MCTS), as a heuristic search algorithm, has achieved superhuman performance in complex domains such as Go [27], chess [28], [29], and real-time strategy games [30].”
Convolutional Neural Networks for Chess Move Prediction with an Emphasis on Top-k Accuracy
Moss, W., & Gandy, L. M. (2025) · Lecture notes in networks and systems · doi:10.1007/978-3-032-07109-5_30
Personalizing Dialog Policy in Health Guidance via Model-based Multitask Reinforcement Learning
Sano, T., Kohjima, M., Takahashi, M., Fujimura, K., Sato, T., & Katagiri, Y. (2025) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn64981.2025.11228619
Adaptive and Scalable Reinforcement Learning Method for Multi-Site Transportation Optimization
Islam, A., & Li, C. (2025) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn64981.2025.11228120
“The methodology has proven equally effective in strategic decision-making systems such as Atari [10], Go [11], and StarCraft II [13], where agents operate in dynamic partially observable scenarios requiring multi-step planning.”
Mixed Policy-Space Response Oracles
Zhang, J., Yang, F., Wang, J., et al. (2025) · doi:10.1109/ijcnn64981.2025.11228019
Grid-Mapping Pseudo-Count Constraint for Offline Reinforcement Learning
Shen, Y., & Huang, H. (2025) · Machine Learning · doi:10.1007/s10994-025-06925-8
“It is worth noting that the vast majority of them do not require interaction with the physical environment.”
A cascading policy learning framework for enhancing power grid resilience
Bensalah, A. S., Marir, T., Yaich, R., & Chebout, M. S. (2025) · PeerJ Computer Science · doi:10.7717/peerj-cs.3358
A Novel CNN-Based Approach for Accurate Binary Classification of Diabetes Using Clinical Data
Mannepalli, P. K., Kaur, K., Singh, P. K., & Kumar, S. (2025) · International Conference on Next Generation of Green Information and Emerging Technologies (GIET) · doi:10.1109/giet65294.2025.11234814
“Causality and explainability are necessary for ML and AI techniques to function similarly to humans, as many of the most successful methods are also the least transparent [20].”
Interdisciplinary Collaboration for AI Development
Hirosawa, T. (2025) · Springer eBooks · doi:10.1007/978-981-95-4338-0_12
Evolving MCTS Macro-actions in Real-Time Domains
Gourley, A., Dann, M., Li, X., & Zambetta, F. (2025) · Lecture notes in computer science · doi:10.1007/978-981-95-4972-6_29
Xia, G., Yang, X., Qi, Q., et al. (2025) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros60139.2025.11247653
Unsupervised classification of non-Hermitian topological phases under symmetries
Long, Y., Xue, H., & Zhang, B. (2025) · National Science Review · doi:10.1093/nsr/nwaf536
AlphaViT: a flexible game-playing AI for multiple games and variable board sizes
Fujita, K. (2025) · PeerJ Computer Science · doi:10.7717/peerj-cs.3403
Reinforcement Learning in Robotics: Unlocking Applications and Advancements
Ahmed, M. F., Tulluri, A. K., & Tiruveedula, N. B. P. (2025) · BENTHAM SCIENCE PUBLISHERS eBooks · doi:10.2174/9789815322316125010011
Generative-AI-assisted knowledge-based augmented exploration in nanophotonics
Iwanaga, M., & Watanabe, K. (2025) · Science and Technology of Advanced Materials Methods · doi:10.1080/27660400.2025.2596940 · Cited by 3
Pradana, A., & Chew, L. Y. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5828847
Gunner: playing FPS game Doom with scalable deep reinforcement learning
Sun, Y., & Zhang, L. (2025) · PeerJ Computer Science · doi:10.7717/peerj-cs.3410
Dang, T., Tran, H., Howard, B., Charoenphon, S., & Nguyen, D. T. (2025) · AI Matters · doi:10.1145/3774399.3774403
“…demon strate that general reinforcement learning algorithms using self-play and extensive training enable AI to challenge humans in Chess and Shogi (Silver et al., 2018), while Brown and Sandholm developed Pluribus, an AI model that achieved superhuman performance in six player no-limit Texas hold…”
InfluentialHahn, N., Moure, P., & Liu, S. (2025) · Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) · doi:10.1109/embc58623.2025.11253749
“In the past AI has demonstrated superhuman performance in complex tasks across various domains, including robotics [13] and games such as Go [30].”
Boosting Indoor Localization Accuracy Using Reinforcement Learning Post-Processing
Adjedjou, H., Ahriz, I., Hamdad, L., & Merhej, D. (2025) · IEEE Annual Ubiquitous Computing, Electronics & Mobile Communication Conference (UEMCON) · doi:10.1109/uemcon67449.2025.11267536
“It is widely recognized for its ability to address sequential decision-making problems in diverse fields such as robotics, games (e.g., AlphaGo [11]), autonomous driving, and finance.”
SRL: A segmented reinforcement learning framework for long sequence layout decisions
Yang, J., Chen, J., Hai, J., Qiao, K., Zhang, H., & Yan, B. (2025) · Complex & Intelligent Systems · doi:10.1007/s40747-025-02193-0
Deep Reinforcement Learning for Equilibrium Computation in Multistage Auctions and Contests
Pieroth, F. R., Kohring, N., & Bichler, M. (2025) · Management Science · doi:10.1287/mnsc.2024.06771 · Cited by 4
“Multi-agent RL has achieved super-human performance in finite games like chess, shogi, Go (Silver et al., 2016; 2018), and imperfect information games like poker (Brown & Sandholm, 2018; 2019).”
Miao, C., Jiang, Y., Cui, Y., et al. (2025) · Applied Intelligence · doi:10.1007/s10489-025-07035-7
Applied Reinforcement Learning: An Implementation Roadmap
Laflamme, C., Viharos, Z. J., Pálvölgyi, B., Schaefer, J., & Haeussler, S. (2025) · IFAC-PapersOnLine · doi:10.1016/j.ifacol.2025.12.024
Rotation-Invariant Game State Evaluation via Board Tensor Canonicalisation
Neshatian, K. (2025) · International Conference on Image and Vision Computing New Zealand (IVCNZ) · doi:10.1109/ivcnz67716.2025.11281816
“The integration of deep learning for policy and value prediction has transformed performance in Go, Chess, and Shogi [2], [3].”
人地系统建模与决策的通用框架初探:多主体深度强化学习及其在公地悲剧中的应用
原, 世., & 李, 新. (2025) · SCIENTIA SINICA Terrae · doi:10.1360/n072025-0092
A Hybrid Minimax-MCTS Chess Engine Enhanced by RLHF Using Stockfish and AlphaZero
Parmar, H., Panthy, S. Y. R., & Murari, U. K. (2025) · IEEE International Conference on Advances in Computing Research on Science Engineering and Technology (ACROSET) · doi:10.1109/acroset66531.2025.11280596 · Cited by 1
“Over the decades, the field has progressed from rudimentary theoretical models to highly sophisticated systems that integrate classical search techniques with modern deep learning frameworks [7], [8], [12].”
InfluentialExplainable physics-based constraints on reinforcement learning for accelerator optimization
Colen, J., Schram, M., Rajput, K., & Kasparian, A. (2025) · Machine Learning Science and Technology · doi:10.1088/2632-2153/ae2fa8 · Cited by 4
“Data-driven techniques such as reinforcement learning (RL) [1] are promising tools to address these challenges, with demonstrated success at playing games [2, 3], solving engineering problems [4, 5] and even scientific tasks [6–9].”
Wermelinger, J., Parduzi, Q., Sariyar, M., Raabe, A., Schneider, U. C., & Seidel, K. (2024) · Figshare · doi:10.6084/m9.figshare.24234312
Prestes, J. A. D. L. (2026) · AI and Ethics · doi:10.1007/s43681-026-01392-w
“Reinforcement learning systems such as AlphaZero provide an early illustration of strategic, future-sensitive action selection [83 ].”
On deep CFR integrated with opponent model in imperfect information games
Wang, J., Li, Y., Niu, S., et al. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.114105
Wermelinger, J., Parduzi, Q., Sariyar, M., Raabe, A., Schneider, U. C., & Seidel, K. (2023) · Figshare · doi:10.6084/m9.figshare.24234247
Reward Shaping for User Satisfaction
Christakopoulou, K., Xu, C., Zhang, S., Badam, S., Potter, T., Li, D., Wan, H., Yi, X., Le, Y., Berg, C., et al. (2025) · Lecture notes in computer science · doi:10.1007/978-3-032-06106-5_4
In Search of Insight: My Life as an Architectural Explorer
Rosenbloom, P. S. (2025) · Journal of Artificial General Intelligence · doi:10.2478/jagi-2025-0002
Liu, Y., Liu, S., Tang, H., Ma, Y., Li, Z., Zhang, J., Xiang, L., & He, Z. (2025) · Knowledge-Based Systems · doi:10.1016/j.knosys.2025.115046 · Cited by 1
Aithal, P. S. (2025) · Poornaprajna International Journal of Teaching & Research Case Studies (PIJTRCS) · doi:10.64818/pijtrcs.3107.8494.0020 · Cited by 3
Wortmann, T. (2025) · Mathematics and the built environment · doi:10.1007/978-3-031-81623-9_22
Kasibhatla, A. S. (2025) · International Journal of Latest Technology in Engineering Management & Applied Science · doi:10.51583/ijltemas.2025.1411000088
Drafting Magic: The Gathering With High-Dimensional Card Embeddings
Bertram, T., Furnkranz, J., & Muller, M. (2025) · IEEE Transactions on Games · doi:10.1109/tg.2025.3646029
“Whilesuperhumanperformancehasbeenachievedin games,suchas Go , chess , DOTA, and Poker [1], [7], [20], many others still pose significant challenges.”
Embracing AI and formalization: Experimenting with tomorrow’s mathematical tools
Alper, J. (2025) · Bulletin of the American Mathematical Society · doi:10.1090/bull/1879 · Cited by 1
“Reinforced learning algorithms such as AlphaZero, after only being taught the rules, have mastered games like chess and Go overnight, becoming far superior than any human player [SHS + 18].”
Wang, H., Du, W., Xu, Y., et al. (2025) · doi:10.1145/3772429.3772436
The impact of AI use on the performance of chess engines
Król, J., & Smółka, J. (2025) · Journal of Computer Sciences Institute · doi:10.35784/jcsi.7972
Multi-Agent Deep Reinforcement Learning for Cyber Conflict Simulation and Experimentation
Corbett, M., Murphy, J. M., & Bastian, N. D. (2025) · IEEE Military Communications Conference (MILCOM) · doi:10.1109/milcom64451.2025.11310735
Large Language Models-Driven Personalized Adaptation Framework for Intelligent Agents
Xu, Z., Fu, C., Sun, N., Huang, H., He, B., & Zhang, X. (2026) · Communications in computer and information science · doi:10.1007/978-981-95-2581-2_27
Metabolic Engineering and Novel Therapeutics
Soni, V., & Upadhyay, A. (2026) · Springer eBooks · doi:10.1007/978-3-031-46895-7_10
Comparison of LLM Models of AI: A Comprehensive Analysis
Kotak, D., Barge, Y., Patel, T., Pandya, N. R., & Adhvarvyu, R. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-06688-6_9
Robotic Perception and Intelligence
Yan, W. Q. (2026) · Advances in computer vision and pattern recognition · doi:10.1007/978-981-95-4360-1_6
Ghita, B., & Zineb, R. (2026) · Lecture notes in networks and systems · doi:10.1007/978-3-032-07785-1_17
Syahbana, Y. A., Esgs, S. P., Wahyudi, M., et al. (2025) · Journal of Innovation Information Technology and Application (JINITA) · doi:10.35970/jinita.v7i2.2691
Kortus, T., Keidel, R., Gauger, N. R., & Kieseler, J. (2026) · Machine Learning Science and Technology · doi:10.1088/2632-2153/ae352b · Cited by 8
“…and optimizing autonomous agents that can compete in a wide variety of complex sequential decision problems, such as games (Mnih et al 2013, Silver et al 2018), robotics (Gu et al 2017, Andrychowicz et al 2020) or autonomous driving (Kendall et al 2019) by discovering complex interaction…”
Song, H. (2025) · Journal of Digital Contents Society · doi:10.9728/dcs.2025.26.12.3533
AI-driven client management in personal services: case study from the beauty industry
Pereshliuga, K. (2025) · InterConf · doi:10.51582/interconf.19-20.10.2025.002
“Clients deserve to know what information is gathered, why it matters, and how it improves their experience [5, p.1142].”
POR UMA REVISÃO CRÍTICA DO USO DE INTELIGÊNCIA ARTIFICIAL NA EDUCAÇÃO
de Lima-Lopes, R. E. (2025) · Revista de Estudos de Cultura · doi:10.32748/revec.v11i27.22719 · Cited by 7
“…DeepMind utilizou a mesma tecnologia de treinamento para construção de outros sistemas, como a aplicação em jogos de tabuleiro como shogi e xadrez (Silver et al. , 2018), predição e mapeamento de estrutura de proteínas (Jumper et al. , 2021) e melhoria de algorítimos complexos de treinamento…”
Hasmukhbhai, C. P., & Khanna, R. (2026) · Smart innovation, systems and technologies · doi:10.1007/978-3-032-12999-4_38
213Chapter 10 AI-driven advances in lung cancer care
Malviya, R., Rajput, S., Roy, M., Sridhar, S. B., & Thomas, S. (2025) · De Gruyter eBooks · doi:10.1515/9783111670744-010
Amini, E., Baniesmaeil, M., Mehdipour, H., Neshat, M., & Marsooli, R. (2026) · Earth-Science Reviews · doi:10.1016/j.earscirev.2026.105399 · Cited by 1
The EGI Framework: A Five-Layer Multi-Agent Cognitive Architecture —Enabling Semantic World Modeling
Wang, C. (2026) · International Journal of Intelligence Science · doi:10.4236/ijis.2026.161007
Energy and spectrum efficient DRL-based algorithms to support sustainable IoT devices
Mazhar, N., Ullah, S. A., & Hassan, S. A. (2026) · Elsevier eBooks · doi:10.1016/b978-0-44-333000-1.00012-2
Zhang, L., Jiang, L., Wu, T., et al. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6122752
HTPCCS: A Hybrid Two-Phase Learning Framework for Congestion Control Scheme Selection
Peng, Z., Zhan, W., Li, B., Chen, X., & Liu, P. (2025) · doi:10.1109/bigdatase66491.2025.00009
“Given the significant achievements of learning-based approaches across numerous domains, e.g., games [3] and networking [4], recent advances in learning-based congestion control have demonstrated the potential of machine learning to improve adaptability in such heterogeneous environments [5]–[11].”
Advancement and challenges of reinforcement learning in lung cancer imaging
Jim, J. R., Hannan, N., Talukder, M. A. R., Mridha, M. F., & Kabir, M. (2026) · Informatics in Medicine Unlocked · doi:10.1016/j.imu.2026.101740 · Cited by 1
Zhang, X., Xu, Y., & Dou, Y. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6007494 · Cited by 1
Xu, R., Liu, X., Zhang, T., et al. (2026) · Journal of Energy Storage · doi:10.1016/j.est.2026.120722
Robust Reinforcement Learning with Dynamic Distortion Risk Measures
Coache, A., & Sebastian Jaimungal, A. (2026) · SIAM Journal on Mathematics of Data Science · doi:10.1137/24m1699802 · Cited by 2
The Authoritative Synthesis Trap: Human-AI Collaboration and the Quality of SciencePreprint
Leippold, M. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.6080627
Baba Is LLM: Reasoning in a Game with Dynamic Rules
van Wetten, F., Plaat, A., & van Duijn, M. J. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-15632-7_4 · Cited by 6
Can AlphaZero Master Human Concepts in Tic-Tac-Toe?
Marasco, A. J., & Givigi, S. N. (2025) · Conference proceedings/Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics · doi:10.1109/smc58881.2025.11343571
Zeng, W., Li, Y., Chen, X., et al. (2025) · Conference proceedings/Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics · doi:10.1109/smc58881.2025.11343206
“This approach, commonly referred to as RL+Search, has been particularly successful in perfect-information games, with AlphaZero [4] representing the pinnacle of this paradigm.”
GNN-ReBeL: Enhancing Neural Belief Representations for Imperfect-Information Games
Zeng, W., Li, Y., Chen, X., et al. (2025) · Conference proceedings/Conference proceedings - IEEE International Conference on Systems, Man, and Cybernetics · doi:10.1109/smc58881.2025.11342940
“The culmination of this line of research was AlphaZero [1], which generalized the approach to multiple games including chess and shogi without domain-specific knowledge beyond the basic game rules.”
Reward Design for Reinforcement Learning in the Development of Large Language Models
Nguyen, T. T., Wilson, C., & Dalins, J. (2025) · International Conference on Advanced Computing and Analytics (ACOMPA) · doi:10.1109/acompa69200.2025.00009
“…AI research, RL gained widespread recognition for its impressive achievements in solving complex problems, with milestones like DeepMind’s AlphaGo [1] and AlphaZero [2] serving as prime examples of RL agents achieving performance at or above human-level in numerous complex games like Go and chess.”
Mahmoud, N., Elaziz, M. A., Dahou, A., Ghatasheh, M., Fares, I. A., Al‐Betar, M. A., & Ewees, A. A. (2026) · Frontiers in Artificial Intelligence · doi:10.3389/frai.2025.1731997
Artificial Intelligence Breakthroughs and Data Futures: A Retrospective and Prospective Review
Yuksel, B. B., & Metin, A. Y. (2026) · Academic Platform Journal of Engineering and Smart Systems · doi:10.21541/apjess.1705042 · Cited by 5
Taylor, P. (2026) · Frontiers in Computer Science · doi:10.3389/fcomp.2025.1639677 · Cited by 1
Checkmating One, by Using Many: Combining Mixture of Experts With MCTS to Improve in Chess
Helfenstein, F., Czech, J., Blüml, J., Max, E., & Kersting, K. (2026) · IEEE Transactions on Games · doi:10.1109/tg.2026.3659995 · Cited by 9
“Despite clear structural differences between the opening, middlegame, and endgame, state-of-the-art engines, such as Al-phaZero [22], MuZero [20], and Stockfish NNUE [16], continue to rely on monolithic architectures trained uniformly across all game phases.”
InfluentialGiving AI agents a sense of control facilitates reinforcement learning in multitasking scenarios
Österdiekhoff, A., Heinrich, N. W., Rußwinkel, N., & Kopp, S. (2026) · PLoS ONE · doi:10.1371/journal.pone.0342305 · Cited by 1
Capaldo, P., Ketheeswaran, V., Quintana-Amate, S., et al. (2026) · Communications in computer and information science · doi:10.1007/978-3-032-15638-9_4
Du, Y. (2026) · Studies in applied philosophy, epistemology and rational ethics · doi:10.1007/978-3-032-15842-0_2
Li, X., Huang, X., Luo, S., Zhou, Q., & Li, Z. (2026) · Expert Systems with Applications · doi:10.1016/j.eswa.2026.131574
(2026) · Elsevier eBooks · doi:10.1016/b978-0-44-327410-7.00040-5
Adaptive Modular Agent Architecture for Hybrid Two-Level Reasoning
Вадимович, Г. Д., Álvarez-Napagao, S., Padget, J., & Ulises Cortés, A. (2026) · Lecture notes in computer science · doi:10.1007/978-3-032-18011-7_2
Artificial intelligence in 6G ecosystem
Gonzalez, H. A., Acevedo, J., Nazeer, K. K., Fournier, C., Aslam, A. R., Huang, J., Lohrmann, M. A., Tietze, R. A., Eichhorn, C., Gumhold, S., et al. (2026) · Elsevier eBooks · doi:10.1016/b978-0-44-327410-7.00024-7
Rokach, Y., & Shapira, B. (2026) · Applied Intelligence · doi:10.1007/s10489-026-07131-2
New Approaches to the Approximation of Solutions in Machine Learning
Gorobtsov, A. S., Рыжов, Е. Н., Орлова, Ю. А., & Donsckaia, A. (2025) · Scientific and Technical Information Processing · doi:10.3103/s014768822570042x
Robotic arm visual-servoing for AI-driven chess gameplay using LLM-based plannner
Ilgun, A. E., Cadirci, O., Ozkan, B., Akgul, B., & Bayraktar, E. (2026) · Neural Computing and Applications · doi:10.1007/s00521-025-11700-w
Generalized Self-Play Reinforcement Learning for Othello under Dynamic Board Constraints
김병창, Lee, Y. L., & Hwang, E. (2025) · International Conference on Information and Communication Technology Convergence (ICTC) · doi:10.1109/ictc66702.2025.11388216
“AlphaZero [1] demonstrated that reinforcement learning agents can attain expert-level performance in games such as Chess, Shogi, and Go through self-play, without the need for expert demonstrations or handcrafted rules.”
Baier, C., Rivera, D. C., Dubslaff, C., Kiebel, S. J., & Wienhöft, P. (2026) · Elsevier eBooks · doi:10.1016/b978-0-44-330044-8.00026-9
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.