{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:2QYOEIHDIFSNFIS6UZR4PC26UG","short_pith_number":"pith:2QYOEIHD","schema_version":"1.0","canonical_sha256":"d430e220e34164d2a25ea663c78b5ea189c16e603365706d6880710b5b7eda65","source":{"kind":"arxiv","id":"2301.02328","version":2},"attestation_state":"computed","paper":{"title":"Extreme Q-Learning: MaxEnt RL without Entropy","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.RO"],"primary_cat":"cs.LG","authors_text":"Divyansh Garg, Joey Hejna, Matthieu Geist, Stefano Ermon","submitted_at":"2023-01-05T23:14:38Z","abstract_excerpt":"Modern Deep Reinforcement Learning (RL) algorithms require estimates of the maximal Q-value, which are difficult to compute in continuous domains with an infinite number of possible actions. In this work, we introduce a new update rule for online and offline RL which directly models the maximal value using Extreme Value Theory (EVT), drawing inspiration from economics. By doing so, we avoid computing Q-values using out-of-distribution actions which is often a substantial source of error. Our key insight is to introduce an objective that directly estimates the optimal soft-value functions (LogS"},"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":"2301.02328","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-01-05T23:14:38Z","cross_cats_sorted":["cs.AI","cs.RO"],"title_canon_sha256":"0edcebd3b4e6dcfa779c9358cd3382ab88a4861f719a66ee098bc3a7fb26acfa","abstract_canon_sha256":"ddaa858d4b4fd532354e164c19d0dd9d1f0c161911168d997f463e07512725f0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:46:53.098109Z","signature_b64":"+ZbN9Lid6YmdRKdNczIqC9EbTGwpQ/R+JRnEdwyeK5WLeJ12uWjRp7ZYrSRhGHG6jm5fK2GCGZFElS7Mw0vwAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d430e220e34164d2a25ea663c78b5ea189c16e603365706d6880710b5b7eda65","last_reissued_at":"2026-07-05T05:46:53.097679Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:46:53.097679Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Extreme Q-Learning: MaxEnt RL without Entropy","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.RO"],"primary_cat":"cs.LG","authors_text":"Divyansh Garg, Joey Hejna, Matthieu Geist, Stefano Ermon","submitted_at":"2023-01-05T23:14:38Z","abstract_excerpt":"Modern Deep Reinforcement Learning (RL) algorithms require estimates of the maximal Q-value, which are difficult to compute in continuous domains with an infinite number of possible actions. In this work, we introduce a new update rule for online and offline RL which directly models the maximal value using Extreme Value Theory (EVT), drawing inspiration from economics. By doing so, we avoid computing Q-values using out-of-distribution actions which is often a substantial source of error. Our key insight is to introduce an objective that directly estimates the optimal soft-value functions (LogS"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2301.02328","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/2301.02328/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":"2301.02328","created_at":"2026-07-05T05:46:53.097737+00:00"},{"alias_kind":"arxiv_version","alias_value":"2301.02328v2","created_at":"2026-07-05T05:46:53.097737+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2301.02328","created_at":"2026-07-05T05:46:53.097737+00:00"},{"alias_kind":"pith_short_12","alias_value":"2QYOEIHDIFSN","created_at":"2026-07-05T05:46:53.097737+00:00"},{"alias_kind":"pith_short_16","alias_value":"2QYOEIHDIFSNFIS6","created_at":"2026-07-05T05:46:53.097737+00:00"},{"alias_kind":"pith_short_8","alias_value":"2QYOEIHD","created_at":"2026-07-05T05:46:53.097737+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11087","citing_title":"Test-Time Gradient Guidance of Flow Policies in Reinforcement Learning","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14779","citing_title":"Peng's Q($\\lambda$) for Conservative Value Estimation in Offline Reinforcement Learning","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.20408","citing_title":"Spectral Souping: A Unified Framework for Online Preference Alignment","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2304.10573","citing_title":"IDQL: Implicit Q-Learning as an Actor-Critic Method with Diffusion Policies","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06104","citing_title":"Beyond Autoregressive RTG: Conditioning via Injection Outside Sequential Modeling in Decision Transformer","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01663","citing_title":"Towards Efficient and Expressive Offline RL via Flow-Anchored Noise-conditioned Q-Learning","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17919","citing_title":"Fisher Decorator: Refining Flow Policy via a Local Transport Map","ref_index":49,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2QYOEIHDIFSNFIS6UZR4PC26UG","json":"https://pith.science/pith/2QYOEIHDIFSNFIS6UZR4PC26UG.json","graph_json":"https://pith.science/api/pith-number/2QYOEIHDIFSNFIS6UZR4PC26UG/graph.json","events_json":"https://pith.science/api/pith-number/2QYOEIHDIFSNFIS6UZR4PC26UG/events.json","paper":"https://pith.science/paper/2QYOEIHD"},"agent_actions":{"view_html":"https://pith.science/pith/2QYOEIHDIFSNFIS6UZR4PC26UG","download_json":"https://pith.science/pith/2QYOEIHDIFSNFIS6UZR4PC26UG.json","view_paper":"https://pith.science/paper/2QYOEIHD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2301.02328&json=true","fetch_graph":"https://pith.science/api/pith-number/2QYOEIHDIFSNFIS6UZR4PC26UG/graph.json","fetch_events":"https://pith.science/api/pith-number/2QYOEIHDIFSNFIS6UZR4PC26UG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2QYOEIHDIFSNFIS6UZR4PC26UG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2QYOEIHDIFSNFIS6UZR4PC26UG/action/storage_attestation","attest_author":"https://pith.science/pith/2QYOEIHDIFSNFIS6UZR4PC26UG/action/author_attestation","sign_citation":"https://pith.science/pith/2QYOEIHDIFSNFIS6UZR4PC26UG/action/citation_signature","submit_replication":"https://pith.science/pith/2QYOEIHDIFSNFIS6UZR4PC26UG/action/replication_record"}},"created_at":"2026-07-05T05:46:53.097737+00:00","updated_at":"2026-07-05T05:46:53.097737+00:00"}