{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:ORJTFBLYVE76OWCAFZCFYR3TTW","short_pith_number":"pith:ORJTFBLY","schema_version":"1.0","canonical_sha256":"7453328578a93fe758402e445c47739dbbb50a17c6c41197030b78e423c57a3c","source":{"kind":"arxiv","id":"2502.03349","version":1},"attestation_state":"computed","paper":{"title":"Robust Autonomy Emerges from Self-Play","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","cs.RO"],"primary_cat":"cs.LG","authors_text":"Aleksei Petrenko, Alex Hertzberg, Brody Huval, David Hafner, Erik Wijmans, Eugene Vinitsky, Marco Cusumano-Towner, Ozan Sener, Philipp Kr\\\"ahenb\\\"uhl, Stuart Bowers, Taylor Killian, Vladlen Koltun","submitted_at":"2025-02-05T16:41:05Z","abstract_excerpt":"Self-play has powered breakthroughs in two-player and multi-player games. Here we show that self-play is a surprisingly effective strategy in another domain. We show that robust and naturalistic driving emerges entirely from self-play in simulation at unprecedented scale -- 1.6~billion~km of driving. This is enabled by Gigaflow, a batched simulator that can synthesize and train on 42 years of subjective driving experience per hour on a single 8-GPU node. The resulting policy achieves state-of-the-art performance on three independent autonomous driving benchmarks. The policy outperforms the pri"},"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":"2502.03349","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-05T16:41:05Z","cross_cats_sorted":["cs.AI","cs.RO"],"title_canon_sha256":"3ac73b4faba42e45adf4a9bd18c4ba59d994ba24131ea317f198dd153528b7c5","abstract_canon_sha256":"3b3d4ad46dabcd226fd1e10a97bc7daaf7aab6cc2649a246f40769c6d4a89552"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:10:03.684976Z","signature_b64":"vgH+YdadCVJfUK8KMz9kFGan9lqxiROPoz/p+LRwj7vdCpnsfMy2k8rpiYe8MFC+eNgEabeaaoo9eorygS6FAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7453328578a93fe758402e445c47739dbbb50a17c6c41197030b78e423c57a3c","last_reissued_at":"2026-07-05T10:10:03.684468Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:10:03.684468Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Robust Autonomy Emerges from Self-Play","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","cs.RO"],"primary_cat":"cs.LG","authors_text":"Aleksei Petrenko, Alex Hertzberg, Brody Huval, David Hafner, Erik Wijmans, Eugene Vinitsky, Marco Cusumano-Towner, Ozan Sener, Philipp Kr\\\"ahenb\\\"uhl, Stuart Bowers, Taylor Killian, Vladlen Koltun","submitted_at":"2025-02-05T16:41:05Z","abstract_excerpt":"Self-play has powered breakthroughs in two-player and multi-player games. Here we show that self-play is a surprisingly effective strategy in another domain. We show that robust and naturalistic driving emerges entirely from self-play in simulation at unprecedented scale -- 1.6~billion~km of driving. This is enabled by Gigaflow, a batched simulator that can synthesize and train on 42 years of subjective driving experience per hour on a single 8-GPU node. The resulting policy achieves state-of-the-art performance on three independent autonomous driving benchmarks. The policy outperforms the pri"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.03349","kind":"arxiv","version":1},"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/2502.03349/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":"2502.03349","created_at":"2026-07-05T10:10:03.684526+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.03349v1","created_at":"2026-07-05T10:10:03.684526+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.03349","created_at":"2026-07-05T10:10:03.684526+00:00"},{"alias_kind":"pith_short_12","alias_value":"ORJTFBLYVE76","created_at":"2026-07-05T10:10:03.684526+00:00"},{"alias_kind":"pith_short_16","alias_value":"ORJTFBLYVE76OWCA","created_at":"2026-07-05T10:10:03.684526+00:00"},{"alias_kind":"pith_short_8","alias_value":"ORJTFBLY","created_at":"2026-07-05T10:10:03.684526+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":14,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25127","citing_title":"Reward-Conditioned Attention: How Reward Design Shapes What Autonomous Driving Agents See","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2606.17386","citing_title":"TerraTransfer: Learning End-to-End Driving Policies Without Expert Demonstrations","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.19370","citing_title":"Human-like autonomy emerges from self-play and a pinch of human data","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2606.25280","citing_title":"EvoFlock: evolved inverse design of multi-agent motion","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2504.13818","citing_title":"Not All Rollouts are Useful: Down-Sampling Rollouts in LLM Reinforcement Learning","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2506.09981","citing_title":"ReSim: Reliable World Simulation for Autonomous Driving","ref_index":122,"is_internal_anchor":false},{"citing_arxiv_id":"2511.23369","citing_title":"SimScale: Learning to Drive via Real-World Simulation at Scale","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2512.05812","citing_title":"Toward Efficient and Robust Behavior Models for Multi-Agent Driving Simulation","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2512.18662","citing_title":"Pseudo-Expert Regularized Offline RL for End-to-End Autonomous Driving in Photorealistic Closed-Loop Environments","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2603.24155","citing_title":"Goal-Oriented Reactive Simulation for Closed-Loop Trajectory Prediction","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27162","citing_title":"A High-Throughput Compute-Efficient POMDP Hide-And-Seek-Engine (HASE) for Multi-Agent Operations","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09153","citing_title":"Beyond Self-Play: Hierarchical Reasoning for Continuous Motion in Closed-Loop Traffic Simulation","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12857","citing_title":"Artificial Intelligence for Modeling and Simulation of Mixed Automated and Human Traffic","ref_index":110,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08535","citing_title":"Fail2Drive: Benchmarking Closed-Loop Driving Generalization","ref_index":11,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ORJTFBLYVE76OWCAFZCFYR3TTW","json":"https://pith.science/pith/ORJTFBLYVE76OWCAFZCFYR3TTW.json","graph_json":"https://pith.science/api/pith-number/ORJTFBLYVE76OWCAFZCFYR3TTW/graph.json","events_json":"https://pith.science/api/pith-number/ORJTFBLYVE76OWCAFZCFYR3TTW/events.json","paper":"https://pith.science/paper/ORJTFBLY"},"agent_actions":{"view_html":"https://pith.science/pith/ORJTFBLYVE76OWCAFZCFYR3TTW","download_json":"https://pith.science/pith/ORJTFBLYVE76OWCAFZCFYR3TTW.json","view_paper":"https://pith.science/paper/ORJTFBLY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.03349&json=true","fetch_graph":"https://pith.science/api/pith-number/ORJTFBLYVE76OWCAFZCFYR3TTW/graph.json","fetch_events":"https://pith.science/api/pith-number/ORJTFBLYVE76OWCAFZCFYR3TTW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ORJTFBLYVE76OWCAFZCFYR3TTW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ORJTFBLYVE76OWCAFZCFYR3TTW/action/storage_attestation","attest_author":"https://pith.science/pith/ORJTFBLYVE76OWCAFZCFYR3TTW/action/author_attestation","sign_citation":"https://pith.science/pith/ORJTFBLYVE76OWCAFZCFYR3TTW/action/citation_signature","submit_replication":"https://pith.science/pith/ORJTFBLYVE76OWCAFZCFYR3TTW/action/replication_record"}},"created_at":"2026-07-05T10:10:03.684526+00:00","updated_at":"2026-07-05T10:10:03.684526+00:00"}