{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:JV7PLT65FZOHT2L77Z7V3AXHUS","short_pith_number":"pith:JV7PLT65","schema_version":"1.0","canonical_sha256":"4d7ef5cfdd2e5c79e97ffe7f5d82e7a4bf35d277a5601c46bb3a07b2c63428ad","source":{"kind":"arxiv","id":"1909.07528","version":2},"attestation_state":"computed","paper":{"title":"Emergent Tool Use From Multi-Agent Autocurricula","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.MA","stat.ML"],"primary_cat":"cs.LG","authors_text":"Bob McGrew, Bowen Baker, Glenn Powell, Igor Mordatch, Ingmar Kanitscheider, Todor Markov, Yi Wu","submitted_at":"2019-09-17T00:17:02Z","abstract_excerpt":"Through multi-agent competition, the simple objective of hide-and-seek, and standard reinforcement learning algorithms at scale, we find that agents create a self-supervised autocurriculum inducing multiple distinct rounds of emergent strategy, many of which require sophisticated tool use and coordination. We find clear evidence of six emergent phases in agent strategy in our environment, each of which creates a new pressure for the opposing team to adapt; for instance, agents learn to build multi-object shelters using moveable boxes which in turn leads to agents discovering that they can over"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"1909.07528","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2019-09-17T00:17:02Z","cross_cats_sorted":["cs.AI","cs.MA","stat.ML"],"title_canon_sha256":"84aced5e1ff82a06e33e2431ca9820df4615b600c6902b1291687c59e5955f52","abstract_canon_sha256":"04374540de49ca025187c270eb724e6a7092fdef6653745798d748e1364ee279"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T00:39:38.450406Z","signature_b64":"AbTSLCSr4bo7Nlr4BWENxjxufWTeDH2wjScR9wITyxl+JVF0JC3dGj3vavfBxuyAqFzSaI79xlm64wuUTv3FDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4d7ef5cfdd2e5c79e97ffe7f5d82e7a4bf35d277a5601c46bb3a07b2c63428ad","last_reissued_at":"2026-07-05T00:39:38.449989Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T00:39:38.449989Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Emergent Tool Use From Multi-Agent Autocurricula","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.MA","stat.ML"],"primary_cat":"cs.LG","authors_text":"Bob McGrew, Bowen Baker, Glenn Powell, Igor Mordatch, Ingmar Kanitscheider, Todor Markov, Yi Wu","submitted_at":"2019-09-17T00:17:02Z","abstract_excerpt":"Through multi-agent competition, the simple objective of hide-and-seek, and standard reinforcement learning algorithms at scale, we find that agents create a self-supervised autocurriculum inducing multiple distinct rounds of emergent strategy, many of which require sophisticated tool use and coordination. We find clear evidence of six emergent phases in agent strategy in our environment, each of which creates a new pressure for the opposing team to adapt; for instance, agents learn to build multi-object shelters using moveable boxes which in turn leads to agents discovering that they can over"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1909.07528","kind":"arxiv","version":2},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/1909.07528/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"1909.07528","created_at":"2026-07-05T00:39:38.450044+00:00"},{"alias_kind":"arxiv_version","alias_value":"1909.07528v2","created_at":"2026-07-05T00:39:38.450044+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1909.07528","created_at":"2026-07-05T00:39:38.450044+00:00"},{"alias_kind":"pith_short_12","alias_value":"JV7PLT65FZOH","created_at":"2026-07-05T00:39:38.450044+00:00"},{"alias_kind":"pith_short_16","alias_value":"JV7PLT65FZOHT2L7","created_at":"2026-07-05T00:39:38.450044+00:00"},{"alias_kind":"pith_short_8","alias_value":"JV7PLT65","created_at":"2026-07-05T00:39:38.450044+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07460","citing_title":"Social-spatial dependencies for learning visual navigation","ref_index":54,"is_internal_anchor":true},{"citing_arxiv_id":"2503.11926","citing_title":"Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2408.00724","citing_title":"Inference Scaling Laws: An Empirical Analysis of Compute-Optimal Inference for Problem-Solving with Language Models","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2309.16797","citing_title":"Promptbreeder: Self-Referential Self-Improvement Via Prompt Evolution","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2603.06977","citing_title":"NePPO: Near-Potential Policy Optimization for General-Sum Multi-Agent Reinforcement Learning","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JV7PLT65FZOHT2L77Z7V3AXHUS","json":"https://pith.science/pith/JV7PLT65FZOHT2L77Z7V3AXHUS.json","graph_json":"https://pith.science/api/pith-number/JV7PLT65FZOHT2L77Z7V3AXHUS/graph.json","events_json":"https://pith.science/api/pith-number/JV7PLT65FZOHT2L77Z7V3AXHUS/events.json","paper":"https://pith.science/paper/JV7PLT65"},"agent_actions":{"view_html":"https://pith.science/pith/JV7PLT65FZOHT2L77Z7V3AXHUS","download_json":"https://pith.science/pith/JV7PLT65FZOHT2L77Z7V3AXHUS.json","view_paper":"https://pith.science/paper/JV7PLT65","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1909.07528&json=true","fetch_graph":"https://pith.science/api/pith-number/JV7PLT65FZOHT2L77Z7V3AXHUS/graph.json","fetch_events":"https://pith.science/api/pith-number/JV7PLT65FZOHT2L77Z7V3AXHUS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JV7PLT65FZOHT2L77Z7V3AXHUS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JV7PLT65FZOHT2L77Z7V3AXHUS/action/storage_attestation","attest_author":"https://pith.science/pith/JV7PLT65FZOHT2L77Z7V3AXHUS/action/author_attestation","sign_citation":"https://pith.science/pith/JV7PLT65FZOHT2L77Z7V3AXHUS/action/citation_signature","submit_replication":"https://pith.science/pith/JV7PLT65FZOHT2L77Z7V3AXHUS/action/replication_record"}},"created_at":"2026-07-05T00:39:38.450044+00:00","updated_at":"2026-07-05T00:39:38.450044+00:00"}