{"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/natural-language-does-not-emerge-naturally-in","title":"Natural Language Does Not Emerge 'Naturally' in Multi-Agent Dialog","arxiv_id":"1706.08502","date":"2017-06-26","proceeding":null,"authors":["Satwik Kottur","José M. F. Moura","Stefan Lee","Dhruv Batra"],"abstract":"A number of recent works have proposed techniques for end-to-end learning of\ncommunication protocols among cooperative multi-agent populations, and have\nsimultaneously found the emergence of grounded human-interpretable language in\nthe protocols developed by the agents, all learned without any human\nsupervision!\n  In this paper, using a Task and Tell reference game between two agents as a\ntestbed, we present a sequence of 'negative' results culminating in a\n'positive' one -- showing that while most agent-invented languages are\neffective (i.e. achieve near-perfect task rewards), they are decidedly not\ninterpretable or compositional.\n  In essence, we find that natural language does not emerge 'naturally',\ndespite the semblance of ease of natural-language-emergence that one may gather\nfrom recent literature. We discuss how it is possible to coax the invented\nlanguages to become more and more human-like and compositional by increasing\nrestrictions on how two agents may communicate.","url_abs":"http://arxiv.org/abs/1706.08502v3","url_pdf":"http://arxiv.org/pdf/1706.08502v3.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":"natural-language-does-not-emerge-naturally-in","repo_url":"https://github.com/batra-mlp-lab/lang-emerge","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"unanswered"}},{"paper_slug":"natural-language-does-not-emerge-naturally-in","repo_url":"https://github.com/kdexd/lang-emerge-parlai","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"natural-language-does-not-emerge-naturally-in","repo_url":"https://github.com/mcogswell/evolang","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1706.08502","atlas_url":"https://app.syntology.ai/?focus=1706.08502","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}