{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,15]],"date-time":"2026-10-15T06:14:32Z","timestamp":1792044872832,"version":"4.4.0"},"reference-count":99,"publisher":"The Royal Society","issue":"2320","license":[{"start":{"date-parts":[[2026,5,14]],"date-time":"2026-05-14T00:00:00Z","timestamp":1778716800000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2026,5,14]],"date-time":"2026-05-14T00:00:00Z","timestamp":1778716800000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100019201","name":"Honda Research Institute","doi-asserted-by":"crossref","id":[{"id":"10.13039\/100019201","id-type":"DOI","asserted-by":"crossref"}]},{"DOI":"10.13039\/100000006","name":"Office of Naval Research","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100000006","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100020769","name":"Center for Brains, Minds, and Machines, Massachusetts Institute of Technology","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100020769","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100006112","name":"Microsoft Research","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100006112","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100006785","name":"Google","doi-asserted-by":"publisher","id":[{"id":"10.13039\/100006785","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,5,14]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Humans are remarkable in their ability to quickly learn to perform complex tasks. Reinforcement learning (RL) has long been proposed as a model of human learning, and while leading machine RL models have surpassed human expertise at many classic board games and video games, they require vast experience to learn successfully\u2014none of today\u2019s algorithms accounts for humans\u2019 ability to learn so many different tasks so quickly. We study human learning on 90 simple yet challenging video games, showing how people learn most games within a few minutes. To explain this behaviour, we propose a strong form of model-based RL, which we call theory-based RL, because it uses cognitively grounded intuitive theories\u2014rich, abstract, causal representations of objects, agents and their interactions\u2014to explore and model an environment and plan effectively to achieve goals. We instantiate the approach in an agent called EMPA (the exploring, modelling and planning agent). EMPA matches human learning efficiency, generalizing robustly to new game situations and levels, as humans do, and exhibiting similar exploration and learning dynamics. Our work points the way for future efforts to build more detailed behavioural models as well as more human-like learning of complex tasks in artificial intelligence systems.<\/jats:p>\n                  <jats:p>This article is part of the theme issue \u2018World models in natural and artificial intelligence\u2019.<\/jats:p>","DOI":"10.1098\/rsta.2024.0529","type":"journal-article","created":{"date-parts":[[2026,5,14]],"date-time":"2026-05-14T08:06:47Z","timestamp":1778746007000},"source":"Crossref","is-referenced-by-count":3,"title":["Human-level learning of complex novel tasks as theory-based modelling, exploration and planning"],"prefix":"10.1098","volume":"384","author":[{"given":"Pedro","family":"Tsividis","sequence":"first","affiliation":[{"name":"MIT 1 , , ,","place":["Cambridge, MA, USA"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Joao","family":"Loula","sequence":"additional","affiliation":[{"name":"MIT 1 , , ,","place":["Cambridge, MA, USA"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jake","family":"Burga","sequence":"additional","affiliation":[{"name":"MIT 1 , , ,","place":["Cambridge, MA, USA"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Juan Pablo","family":"Rodriguez","sequence":"additional","affiliation":[{"name":"MIT 1 , , ,","place":["Cambridge, MA, USA"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sergio","family":"Arnaud","sequence":"additional","affiliation":[{"name":"Deep Dive AI 2 , ,","place":["Mexico City, Mexico"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nate","family":"Foss","sequence":"additional","affiliation":[{"name":"MIT 1 , , ,","place":["Cambridge, MA, USA"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Andres","family":"Campero","sequence":"additional","affiliation":[{"name":"MIT 1 , , ,","place":["Cambridge, MA, USA"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ajay","family":"Subramanian","sequence":"additional","affiliation":[{"name":"NYU 3 , , ,","place":["New York, NY, USA"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Thomas","family":"Pouncy","sequence":"additional","affiliation":[{"name":"Harvard University 4 , , ,","place":["Cambridge, MA, USA"],"department":["Department of Psychology"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-6546-3298","authenticated-orcid":false,"given":"Samuel J.","family":"Gershman","sequence":"additional","affiliation":[{"name":"Harvard University 4 , , ,","place":["Cambridge, MA, USA"],"department":["Department of Psychology"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0138-163X","authenticated-orcid":false,"given":"Joshua B.","family":"Tenenbaum","sequence":"additional","affiliation":[{"name":"MIT 1 , , ,","place":["Cambridge, MA, USA"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"175","published-online":{"date-parts":[[2026,5,14]]},"reference":[{"key":"2026051406034237400_R1","doi-asserted-by":"crossref","first-page":"135","DOI":"10.1037\/0033-295X.88.2.135","article-title":"Toward a modern theory of adaptive networks: expectation and prediction","volume":"88","author":"Sutton","year":"1981","journal-title":"Psychol. Rev."},{"key":"2026051406034237400_R2","doi-asserted-by":"crossref","first-page":"1593","DOI":"10.1126\/science.275.5306.1593","article-title":"A neural substrate of prediction and reward","volume":"275","author":"Schultz","year":"1997","journal-title":"Science"},{"key":"2026051406034237400_R3","first-page":"279","article-title":"Q-learning","volume":"8","author":"Watkins","year":"1992","journal-title":"Mach. Learn."},{"key":"2026051406034237400_R4","doi-asserted-by":"crossref","first-page":"876","DOI":"10.1038\/nature04766","article-title":"Cortical substrates for exploratory decisions in humans","volume":"441","author":"Daw","year":"2006","journal-title":"Nature"},{"key":"2026051406034237400_R5","unstructured":"Guo\n              X\n            , SinghS, LeeH, LewisRL, WangX. 2014Deep learning for real-time Atari game play using offline Monte-Carlo tree search planning. InAdvances in neural information processing systems, pp. 3338\u20133346. La Jolla, CA: NeurIPS Foundation."},{"key":"2026051406034237400_R6","doi-asserted-by":"crossref","first-page":"529","DOI":"10.1038\/nature14236","article-title":"Human-level control through deep reinforcement learning","volume":"518","author":"Mnih","year":"2015","journal-title":"Nature"},{"key":"2026051406034237400_R7","doi-asserted-by":"crossref","DOI":"10.1609\/aaai.v30i1.10295","article-title":"Deep reinforcement learning with double q-learning","volume":"30","author":"Van Hasselt","year":"2016","journal-title":"AAAI"},{"key":"2026051406034237400_R8","article-title":"Prioritized experience replay","author":"Schaul","year":"2015","journal-title":"arXiv Preprint arXiv:1511.05952"},{"key":"2026051406034237400_R9","article-title":"Incentivizing exploration in reinforcement learning with deep predictive models","author":"Stadie","year":"2015","journal-title":"arXiv Preprint arXiv:1507.00814"},{"key":"2026051406034237400_R10","unstructured":"Mnih\n              V\n            , BadiaAP, MirzaM, GravesA, LillicrapT, HarleyT, SilverD, KavukcuogluK. 2016Asynchronous methods for deep reinforcement learning. InInt. Conf. on Machine Learning, pp. 1928\u20131937. Cambridge, MA: PMLR."},{"key":"2026051406034237400_R11","article-title":"Learning to play in a day: faster deep reinforcement learning by optimality tightening","author":"He","year":"2016","journal-title":"arXiv Preprint arXiv:1611.01606"},{"key":"2026051406034237400_R12","article-title":"Rainbow: combining improvements in deep reinforcement learning","author":"Hessel","year":"2017","journal-title":"arXiv Preprint arXiv:1710.02298"},{"key":"2026051406034237400_R13","unstructured":"Tsividis\n              P\n            , PouncyT, XuJ, TenenbaumJB, GershmanSJ. 2017Human learning in Atari. InAAAI spring symposium series. Palo Alto, CA: AAAI Press."},{"key":"2026051406034237400_R14","unstructured":"Kansky\n              K\n            \n            et al\n          . 2017Schema networks: zero-shot transfer with a generative causal model of intuitive physics. InInt. Conf. on Machine Learning, pp. 1809\u20131818. Cambridge, MA: PMLR."},{"key":"2026051406034237400_R15","volume-title":"Reinforcement learning: an introduction","author":"Sutton","year":"2018"},{"key":"2026051406034237400_R16","doi-asserted-by":"crossref","first-page":"29","DOI":"10.1207\/s15516709cog1401_3","article-title":"Principles of object perception","volume":"14","author":"Spelke","year":"1990","journal-title":"Cogn. Sci."},{"key":"2026051406034237400_R17","doi-asserted-by":"crossref","first-page":"89","DOI":"10.1111\/j.0963-7214.2004.00281.x","article-title":"Infants\u2019 physical world","volume":"13","author":"Baillargeon","year":"2004","journal-title":"Curr. Dir. Psychol. Sci."},{"key":"2026051406034237400_R18","doi-asserted-by":"crossref","first-page":"89","DOI":"10.1111\/j.1467-7687.2007.00569.x","article-title":"Core knowledge","volume":"10","author":"Spelke","year":"2007","journal-title":"Dev. Sci."},{"key":"2026051406034237400_R19","doi-asserted-by":"crossref","first-page":"705","DOI":"10.1016\/j.cognition.2007.08.001","article-title":"Goal attribution to inanimate agents by 6.5-month-old infants","volume":"107","author":"Csibra","year":"2008","journal-title":"Cognition"},{"key":"2026051406034237400_R20","doi-asserted-by":"crossref","first-page":"3","DOI":"10.1037\/0033-295X.111.1.3","article-title":"A theory of causal learning in children: causal maps and Bayes nets","volume":"111","author":"Gopnik","year":"2004","journal-title":"Psychol. Rev."},{"key":"2026051406034237400_R21","doi-asserted-by":"crossref","first-page":"289","DOI":"10.1037\/0033-295X.92.3.289","article-title":"The role of theories in conceptual coherence","volume":"92","author":"Murphy","year":"1985","journal-title":"Psychol. Rev."},{"key":"2026051406034237400_R22","volume-title":"Conceptual change in childhood","author":"Carey","year":"1985"},{"key":"2026051406034237400_R23","doi-asserted-by":"crossref","DOI":"10.7551\/mitpress\/7289.001.0001","volume-title":"Words, thoughts, and theories","author":"Gopnik","year":"1997"},{"key":"2026051406034237400_R24","doi-asserted-by":"crossref","first-page":"780","DOI":"10.1038\/29528","article-title":"Signal-dependent noise determines motor planning","volume":"394","author":"Harris","year":"1998","journal-title":"Nature"},{"key":"2026051406034237400_R25","doi-asserted-by":"crossref","first-page":"718","DOI":"10.1016\/S0959-4388(99)00028-8","article-title":"Internal models for motor control and trajectory planning","volume":"9","author":"Kawato","year":"1999","journal-title":"Curr. Opin. Neurobiol."},{"key":"2026051406034237400_R26","doi-asserted-by":"crossref","DOI":"10.1038\/srep31378","article-title":"Model-based action planning involves cortico-cerebellar and basal ganglia networks","volume":"6","author":"Fermin","year":"2016","journal-title":"Sci. Rep."},{"key":"2026051406034237400_R27","unstructured":"Guestrin\n              C\n            , KollerD, GearhartC, KanodiaN. 2003Generalizing plans to new environments in relational mdps. InProc. 18th Int. Joint Conf. on Artificial Intelligence, pp. 1003\u20131010. San Francisco, CA: IJCAI Organization."},{"key":"2026051406034237400_R28","unstructured":"Pasula\n              H\n            , ZettlemoyerLS, KaelblingLP. 2004Learning probabilistic relational planning rules. InICAPS, pp. 73\u201382. Palo Alto, CA: AAAI Press."},{"key":"2026051406034237400_R29","doi-asserted-by":"crossref","first-page":"309","DOI":"10.1613\/jair.2113","article-title":"Learning symbolic models of stochastic domains","volume":"29","author":"Pasula","year":"2007","journal-title":"Jair"},{"key":"2026051406034237400_R30","article-title":"Learning sparse relational transition models","author":"Xia","year":"2018","journal-title":"arXiv Preprint arXiv:1810.11177"},{"key":"2026051406034237400_R31","unstructured":"Scholz\n              J\n            , LevihnM, IsbellC, WingateD. 2014A physics-based model prior for object-oriented mdps. InInt. Conf. on Machine Learning, pp. 1089\u20131097. Cambridge, MA: PMLR."},{"key":"2026051406034237400_R32","article-title":"Strategic object oriented reinforcement learning","author":"Keramati","year":"2018","journal-title":"arXiv Preprint arXiv:1806.00175"},{"key":"2026051406034237400_R33","unstructured":"Oh\n              J\n            , GuoX, LeeH, LewisRL, SinghS. 2015Action-conditional video prediction using deep networks in Atari games. InAdvances in Neural Information Processing Systems, pp. 2863\u20132871. La Jolla, CA: NeurIPS Foundation."},{"key":"2026051406034237400_R34","article-title":"Learning visual predictive models of physics for playing billiards","author":"Fragkiadaki","year":"2015","journal-title":"arXiv Preprint arXiv:1511.07404"},{"key":"2026051406034237400_R35","article-title":"Recurrent environment simulators","author":"Chiappa","year":"2017","journal-title":"arXiv Preprint arXiv:1704.02254"},{"key":"2026051406034237400_R36","article-title":"A deep learning approach for joint video frame and reward prediction in atari games","author":"Leibfried","year":"2016","journal-title":"arXiv Preprint arXiv:1611.07078"},{"key":"2026051406034237400_R37","article-title":"World models","author":"Ha","year":"2018","journal-title":"arXiv Preprint arXiv:1803.10122"},{"key":"2026051406034237400_R38","unstructured":"Racani\u00e8re\n              S\n            \n            et al\n          . 2017Imagination-augmented agents for deep reinforcement learning. InAdvances in Neural Information Processing Systems, pp. 5690\u20135701. La Jolla, CA: NeurIPS Foundation."},{"key":"2026051406034237400_R39","article-title":"Model-based reinforcement learning for atari","author":"Kaiser","year":"2019","journal-title":"arXiv Preprint arXiv:1903.00374"},{"key":"2026051406034237400_R40","article-title":"COBRA: data-efficient model-based RL through unsupervised object discovery and curiosity-driven exploration","author":"Watters","year":"2019","journal-title":"arXiv Preprint arXiv:1905.09275"},{"key":"2026051406034237400_R41","doi-asserted-by":"crossref","first-page":"604","DOI":"10.1038\/s41586-020-03051-4","article-title":"Mastering Atari, Go, chess and shogi by planning with a learned model","volume":"588","author":"Schrittwieser","year":"2020","journal-title":"Nature"},{"key":"2026051406034237400_R42","unstructured":"Hafner\n              D\n            , LillicrapTP, NorouziM, BaJ. 2021Mastering Atari with discrete world models. InInt. Conf. on Learning Representations. https:\/\/openreview.net\/forum?id=0oabwyZbOu."},{"key":"2026051406034237400_R43","doi-asserted-by":"crossref","first-page":"1204","DOI":"10.1016\/j.neuron.2011.02.027","article-title":"Model-based influences on humans\u2019 choices and striatal prediction errors","volume":"69","author":"Daw","year":"2011","journal-title":"Neuron"},{"key":"2026051406034237400_R44","doi-asserted-by":"crossref","first-page":"751","DOI":"10.1177\/0956797612463080","article-title":"The curse of planning: dissecting multiple reinforcement-learning systems by taxing the central executive","volume":"24","author":"Otto","year":"2013","journal-title":"Psychol. Sci."},{"key":"2026051406034237400_R45","doi-asserted-by":"crossref","first-page":"1321","DOI":"10.1177\/0956797617708288","article-title":"Cost-benefit arbitration between multiple reinforcement-learning systems","volume":"28","author":"Kool","year":"2017","journal-title":"Psychol. Sci."},{"key":"2026051406034237400_R46","doi-asserted-by":"crossref","DOI":"10.1017\/S0140525X16001837","article-title":"Building machines that learn and think like people","volume":"40","author":"Lake","year":"2017","journal-title":"Behav. Brain Sci."},{"key":"2026051406034237400_R47","doi-asserted-by":"crossref","first-page":"29302","DOI":"10.1073\/pnas.1912341117","article-title":"Rapid trial-and-error learning with simulation supports flexible tool use and physical reasoning","volume":"117","author":"Allen","year":"2020","journal-title":"Proc. Natl Acad. Sci. USA"},{"key":"2026051406034237400_R48","doi-asserted-by":"crossref","first-page":"5012","DOI":"10.1073\/pnas.0704450105","article-title":"Intuitive statistics by 8-month-old infants","volume":"105","author":"Xu","year":"2008","journal-title":"Proc. Natl Acad. Sci. USA"},{"key":"2026051406034237400_R49","doi-asserted-by":"crossref","first-page":"211","DOI":"10.1016\/j.cognition.2008.07.017","article-title":"Going beyond the evidence: abstract laws and preschoolers\u2019 responses to anomalous data","volume":"109","author":"Schulz","year":"2008","journal-title":"Cognition"},{"key":"2026051406034237400_R50","doi-asserted-by":"crossref","first-page":"341","DOI":"10.1016\/j.cognition.2011.03.003","article-title":"Where science starts: spontaneous experiments in preschoolers\u2019 exploratory play","volume":"120","author":"Cook","year":"2011","journal-title":"Cognition"},{"key":"2026051406034237400_R51","unstructured":"Tsividis\n              P\n            , GershmanS, TenenbaumJ, SchulzL. 2014Information selection in noisy environments with large action spaces. InProceedings of the 36th annual conference of the cognitive science society, pp. 1622\u20131627. Quebec City, Canada."},{"key":"2026051406034237400_R52","doi-asserted-by":"crossref","first-page":"661","DOI":"10.1037\/a0017201","article-title":"Theory-based causal induction","volume":"116","author":"Griffiths","year":"2009","journal-title":"Psychol. Rev."},{"key":"2026051406034237400_R53","doi-asserted-by":"crossref","DOI":"10.1093\/oso\/9780190618247.001.0001","volume-title":"What babies know: core knowledge and composition","author":"Spelke","year":"2022"},{"key":"2026051406034237400_R54","volume-title":"Advances in child development and behavior","author":"Schulz","year":"2012"},{"key":"2026051406034237400_R55","unstructured":"Tsividis\n              P\n            , TenenbaumJB, SchulzL. 2015Hypothesis-space constraints in causal learning. InCogSci...Annual Conference of the Cognitive Science Society. Cognitive Science Society (U.S.). Conference. Austin, TX: Cognitive Science Society."},{"key":"2026051406034237400_R56","doi-asserted-by":"crossref","first-page":"229","DOI":"10.1109\/TCIAIG.2015.2402393","article-title":"The 2014 general video game playing competition","volume":"8","author":"Perez-Liebana","year":"2016","journal-title":"IEEE Trans. Comput. Intell. AI Games"},{"key":"2026051406034237400_R57","unstructured":"Schaul\n              T\n            \n          . 2013InConference on Computational Intelligence and Games (CIG), pp. 1\u20138. Niagara Falls, ON, Canada. (doi:10.1109\/CIG.2013.6633610)"},{"key":"2026051406034237400_R58","doi-asserted-by":"crossref","first-page":"986","DOI":"10.1214\/aoms\/1177728069","article-title":"On a measure of the information provided by an experiment","volume":"27","author":"Lindley","year":"1956","journal-title":"Ann. Math. Statist."},{"key":"2026051406034237400_R59","doi-asserted-by":"crossref","first-page":"686","DOI":"10.1214\/aos\/1176344689","article-title":"Expected information as expected utility","volume":"7","author":"Bernardo","year":"1979","journal-title":"Ann. Statist."},{"key":"2026051406034237400_R60","volume-title":"Width and serialization of classical planning problems","author":"Geffner","year":"2012"},{"key":"2026051406034237400_R61","doi-asserted-by":"crossref","first-page":"3590","DOI":"10.1609\/aaai.v31i1.11027","article-title":"Best-first width search: exploration and exploitation in classical planning","volume":"31","author":"Lipovetzky","year":"2017","journal-title":"AAAI"},{"key":"2026051406034237400_R62","article-title":"Noisy networks for exploration","author":"Fortunato","year":"2017","journal-title":"arXiv Preprint arXiv:1706.10295"},{"key":"2026051406034237400_R63","doi-asserted-by":"crossref","first-page":"18327","DOI":"10.1073\/pnas.1306572110","article-title":"Simulation as an engine of physical scene understanding","volume":"110","author":"Battaglia","year":"2013","journal-title":"Proc. Natl Acad. Sci. USA"},{"key":"2026051406034237400_R64","doi-asserted-by":"crossref","first-page":"1731","DOI":"10.1177\/0956797617713053","article-title":"Eye-tracking causality","volume":"28","author":"Gerstenberg","year":"2017","journal-title":"Psychol. Sci."},{"key":"2026051406034237400_R65","article-title":"Mastering Atari games with limited data","volume":"34","author":"Ye","year":"2021","journal-title":"Adv. Neural Inf. Process. Syst."},{"key":"2026051406034237400_R66","doi-asserted-by":"crossref","first-page":"427","DOI":"10.1111\/j.1467-8624.2006.00880.x","article-title":"God does not play dice: causal determinism and preschoolers\u2019 causal inferences","volume":"77","author":"Schulz","year":"2006","journal-title":"Child Dev."},{"key":"2026051406034237400_R67","doi-asserted-by":"crossref","DOI":"10.1093\/acprof:oso\/9780195154061.001.0001","volume-title":"The essential child: origins of essentialism in everyday thought","author":"Gelman","year":"2003"},{"key":"2026051406034237400_R68","doi-asserted-by":"crossref","DOI":"10.1111\/cogs.12928","article-title":"What is the model in model-based planning?","volume":"45","author":"Pouncy","year":"2021","journal-title":"Cogn. Sci."},{"key":"2026051406034237400_R69","doi-asserted-by":"crossref","first-page":"101509","DOI":"10.1016\/j.cogpsych.2022.101509","article-title":"Inductive biases in theory-based reinforcement learning","volume":"138","author":"Pouncy","year":"2022","journal-title":"Cogn. Psychol."},{"key":"2026051406034237400_R70","doi-asserted-by":"crossref","first-page":"1331","DOI":"10.1016\/j.neuron.2023.01.023","article-title":"The neural architecture of theory-based reinforcement learning","volume":"111","author":"Tomov","year":"2023","journal-title":"Neuron"},{"key":"2026051406034237400_R71","doi-asserted-by":"crossref","first-page":"1000","DOI":"10.1038\/s41586-023-06124-2","article-title":"Expertise increases planning depth in human gameplay","volume":"618","author":"van Opheusden","year":"2023","journal-title":"Nature"},{"key":"2026051406034237400_R72","doi-asserted-by":"crossref","first-page":"191","DOI":"10.1613\/jair.1705","article-title":"The fast downward planning system","volume":"26","author":"Helmert","year":"2006","journal-title":"Jair"},{"key":"2026051406034237400_R73","unstructured":"Garriga Alonso\n              A\n            \n          . 2017Solving Montezuma\u2019s revenge with planning and reinforcement learning: Master\u2019s thesis. Universitat Pompeu Fabra, Barcelona."},{"key":"2026051406034237400_R74","doi-asserted-by":"crossref","DOI":"10.1017\/S0140525X17000280","article-title":"Avoiding frostbite: It helps to learn from others","volume":"40","author":"Tessler","year":"2017","journal-title":"Behav. Brain Sci."},{"key":"2026051406034237400_R75","unstructured":"Tessler\n              MH\n            , MadeanoJ, TsividisPA, HarperB, GoodmanND, TenenbaumJB. 2021Learning to solve complex tasks by growing knowledge culturally across generations. Seehttps:\/\/arxiv.org\/abs\/2107.13377."},{"key":"2026051406034237400_R76","doi-asserted-by":"crossref","unstructured":"Diuk\n              C\n            , CohenA, LittmanML. 2008An object-oriented representation for efficient reinforcement learning. In25th Int. Conf. on Machine Learning, Helsinki, Finland, pp. 240\u2013247. New York, NY: ACM. (doi:10.1145\/1390156.1390187)","DOI":"10.1145\/1390156.1390187"},{"key":"2026051406034237400_R77","article-title":"Bayesian reinforcement learning: a survey","author":"Ghavamzadeh","year":"2016","journal-title":"arXiv Preprint arXiv:1609.04436"},{"key":"2026051406034237400_R78","doi-asserted-by":"crossref","unstructured":"Tam\n              A\n            \n            et al\n          . 2022Semantic exploration from language abstractions and pretrained representations. InAdvances in Neural Information Processing Systems 35, vol. 35, pp. 25377\u201325389, San Diego, CA.","DOI":"10.52202\/068431-1840"},{"key":"2026051406034237400_R79","doi-asserted-by":"crossref","unstructured":"Mu\n              J\n            \n            et al\n          . 2022Improving intrinsic exploration with language abstractions. InAdvances in Neural Information Processing Systems 35, vol. 35, pp. 33947\u201333960, San Diego, CA.","DOI":"10.52202\/068431-2460"},{"key":"2026051406034237400_R80","article-title":"Guiding pretraining in reinforcement learning with large language models","author":"Du","year":"2023","journal-title":"arXiv Preprint arXiv:2302.06692"},{"key":"2026051406034237400_R81","doi-asserted-by":"crossref","first-page":"860","DOI":"10.1038\/s41593-018-0147-8","article-title":"Prefrontal cortex as a meta-reinforcement learning system","volume":"21","author":"Wang","year":"2018","journal-title":"Nat. Neurosci."},{"key":"2026051406034237400_R82","article-title":"Impala: scalable distributed deep-rl with importance weighted actor-learner architectures","author":"Espeholt","year":"2018","journal-title":"arXiv Preprint arXiv:1802.01561"},{"key":"2026051406034237400_R83","article-title":"Human-level performance in first-person multiplayer games with population-based deep reinforcement learning","author":"Jaderberg","year":"2018","journal-title":"arXiv Preprint arXiv:1807.01281"},{"key":"2026051406034237400_R84","article-title":"Causal reasoning from meta-reinforcement learning","author":"Dasgupta","year":"2019","journal-title":"arXiv Preprint arXiv:1901.08162"},{"key":"2026051406034237400_R85","article-title":"Evolution strategies as a scalable alternative to reinforcement learning","author":"Salimans","year":"2017","journal-title":"arXiv Preprint arXiv:1703.03864"},{"key":"2026051406034237400_R86","unstructured":"Ellis\n              K\n            , MoralesL, MeyerMS, Solar-LezamaA, TenenbaumJB. 2018Dreamcoder: bootstrapping domain-specific languages for neurally-guided Bayesian program learning. InProc. 2nd Workshop on Neural Abstract Machines and Program Induction. La Jolla, CA: NeurIPS Foundation."},{"key":"2026051406034237400_R87","unstructured":"Das\n              R\n            , TenenbaumJB, Solar-LezamaA, TavaresZ. 2021AutumnSynth: synthesis of reactive programs with structured latent state. InAdvances in programming languages and neurosymbolic systems workshop. La Jolla, CA: NeurIPS Foundation."},{"key":"2026051406034237400_R88","unstructured":"Tang\n              H\n            , KeyD, EllisK. 2024WorldCoder, a model-based LLM agent: building world models by writing code and interacting with the environment. Seehttps:\/\/arxiv.org\/abs\/2402.12275."},{"key":"2026051406034237400_R89","article-title":"Relational inductive biases, deep learning, and graph networks","author":"Battaglia","year":"2018","journal-title":"arXiv Preprint arXiv:1806.01261"},{"key":"2026051406034237400_R90","article-title":"A compositional object-based approach to learning physical dynamics","author":"Chang","year":"2016","journal-title":"arXiv Preprint arXiv:1612.00341"},{"key":"2026051406034237400_R91","unstructured":"Watters\n              N\n            , ZoranD, WeberT, BattagliaP, PascanuR, TacchettiA. 2017Visual interaction networks: learning a physics simulator from video. InAdvances in Neural Information Processing Systems, pp. 4539\u20134547. La Jolla, CA: NeurIPS Foundation."},{"key":"2026051406034237400_R92","article-title":"Relational deep reinforcement learning","author":"Zambaldi","year":"2018","journal-title":"arXiv Preprint arXiv:1806.01830"},{"key":"2026051406034237400_R93","article-title":"Towards deep symbolic reinforcement learning","author":"Garnelo","year":"2016","journal-title":"arXiv Preprint arXiv:1609.05518"},{"key":"2026051406034237400_R94","unstructured":"Heins\n              C\n            \n            et al\n          . 2025AXIOM: learning to play games in minutes with expanding object-centric models. Seehttps:\/\/arxiv.org\/abs\/2505.24784."},{"key":"2026051406034237400_R95","unstructured":"Hafner\n              D\n            , PasukonisJ, BaJ, LillicrapT. 2024Mastering diverse domains through world models. Seehttps:\/\/arxiv.org\/abs\/2301.04104."},{"key":"2026051406034237400_R96","doi-asserted-by":"crossref","DOI":"10.1093\/acprof:oso\/9780195367638.001.0001","volume-title":"The origin of concepts","author":"Carey","year":"2009"},{"key":"2026051406034237400_R97","doi-asserted-by":"crossref","first-page":"1010","DOI":"10.1609\/aaai.v25i1.8013","article-title":"Qualitative numeric planning","volume":"25","author":"Srivastava","year":"2011","journal-title":"AAAI"},{"key":"2026051406034237400_R98","doi-asserted-by":"crossref","first-page":"2703","DOI":"10.1609\/aaai.v33i01.33012703","article-title":"Learning features and abstract actions for computing generalized plans","volume":"33","author":"Bonet","year":"2019","journal-title":"AAAI"},{"key":"2026051406034237400_R99","article-title":"Supplementary material from: Human-level Learning of Complex Tasks Through Theory-Based World Modeling, Exploration, and Planning. Figshare.","author":"Tsividis","year":"2026"}],"container-title":["Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/royalsocietypublishing.org\/rsta\/article-pdf\/doi\/10.1098\/rsta.2024.0529\/6137905\/rsta.2024.0529.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/royalsocietypublishing.org\/rsta\/article-pdf\/doi\/10.1098\/rsta.2024.0529\/6137905\/rsta.2024.0529.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,5,14]],"date-time":"2026-05-14T10:04:30Z","timestamp":1778753070000},"score":1,"resource":{"primary":{"URL":"https:\/\/royalsocietypublishing.org\/rsta\/article\/384\/2320\/20240529\/481693\/Human-level-learning-of-complex-novel-tasks-as"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,14]]},"references-count":99,"journal-issue":{"issue":"2320","published-online":{"date-parts":[[2026,5,14]]},"published-print":{"date-parts":[[2026,5,14]]}},"URL":"https:\/\/doi.org\/10.1098\/rsta.2024.0529","relation":{},"ISSN":["1364-503X","1471-2962"],"issn-type":[{"value":"1364-503X","type":"print"},{"value":"1471-2962","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5,14]]},"article-number":"20240529"}}