{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/emergent-complexity-via-multi-agent","title":"Emergent Complexity via Multi-Agent Competition","arxiv_id":"1710.03748","date":"2017-10-10","proceeding":"ICLR 2018 1","authors":["Trapit Bansal","Jakub Pachocki","Szymon Sidor","Ilya Sutskever","Igor Mordatch"],"abstract":"Reinforcement learning algorithms can train agents that solve problems in\ncomplex, interesting environments. Normally, the complexity of the trained\nagent is closely related to the complexity of the environment. This suggests\nthat a highly capable agent requires a complex environment for training. In\nthis paper, we point out that a competitive multi-agent environment trained\nwith self-play can produce behaviors that are far more complex than the\nenvironment itself. We also point out that such environments come with a\nnatural curriculum, because for any skill level, an environment full of agents\nof this level will have the right level of difficulty. This work introduces\nseveral competitive multi-agent environments where agents compete in a 3D world\nwith simulated physics. The trained agents learn a wide variety of complex and\ninteresting skills, even though the environment themselves are relatively\nsimple. The skills include behaviors such as running, blocking, ducking,\ntackling, fooling opponents, kicking, and defending using both arms and legs. A\nhighlight of the learned behaviors can be found here: https://goo.gl/eR7fbX","url_abs":"http://arxiv.org/abs/1710.03748v3","url_pdf":"http://arxiv.org/pdf/1710.03748v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"emergent-complexity-via-multi-agent","repo_url":"https://github.com/openai/multiagent-competition","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"tf","reach":null},{"paper_slug":"emergent-complexity-via-multi-agent","repo_url":"https://github.com/vgudapati/DRLND_Collaboration_Competetion","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"blocking","task_name":"Blocking"},{"task_slug":"reinforcement-learning","task_name":"Reinforcement Learning"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":"https://app.syntology.ai/?focus=1710.03748","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}