{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:XPJT74A556ZMM7EH2UYWEIAYX4","short_pith_number":"pith:XPJT74A5","schema_version":"1.0","canonical_sha256":"bbd33ff01defb2c67c87d531622018bf3fdfec3bec3998f10fada3ff59e4025f","source":{"kind":"arxiv","id":"2007.14430","version":3},"attestation_state":"computed","paper":{"title":"Munchausen Reinforcement Learning","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Matthieu Geist, Nino Vieillard, Olivier Pietquin","submitted_at":"2020-07-28T18:30:23Z","abstract_excerpt":"Bootstrapping is a core mechanism in Reinforcement Learning (RL). Most algorithms, based on temporal differences, replace the true value of a transiting state by their current estimate of this value. Yet, another estimate could be leveraged to bootstrap RL: the current policy. Our core contribution stands in a very simple idea: adding the scaled log-policy to the immediate reward. We show that slightly modifying Deep Q-Network (DQN) in that way provides an agent that is competitive with distributional methods on Atari games, without making use of distributional RL, n-step returns or prioritize"},"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":"2007.14430","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2020-07-28T18:30:23Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"7ebf7b64dc5758385f96bfb96d71babb8836af7c19c27934e171f9ac90a07c22","abstract_canon_sha256":"463f677d698a949ccaac2dafceb06b583689fe8711dbe606d308919adf76cb1b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:49:04.584977Z","signature_b64":"EnqxHs+jvGUlYa8y/VtZWtpAG1SBN5hm4S6bf7djQJlM8hQlI6ipjZJYyQAHBXOoptEPI+gcoZguZPT7fpGQAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bbd33ff01defb2c67c87d531622018bf3fdfec3bec3998f10fada3ff59e4025f","last_reissued_at":"2026-07-05T01:49:04.584523Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:49:04.584523Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Munchausen Reinforcement Learning","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Matthieu Geist, Nino Vieillard, Olivier Pietquin","submitted_at":"2020-07-28T18:30:23Z","abstract_excerpt":"Bootstrapping is a core mechanism in Reinforcement Learning (RL). Most algorithms, based on temporal differences, replace the true value of a transiting state by their current estimate of this value. Yet, another estimate could be leveraged to bootstrap RL: the current policy. Our core contribution stands in a very simple idea: adding the scaled log-policy to the immediate reward. We show that slightly modifying Deep Q-Network (DQN) in that way provides an agent that is competitive with distributional methods on Atari games, without making use of distributional RL, n-step returns or prioritize"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2007.14430","kind":"arxiv","version":3},"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/2007.14430/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":"2007.14430","created_at":"2026-07-05T01:49:04.584588+00:00"},{"alias_kind":"arxiv_version","alias_value":"2007.14430v3","created_at":"2026-07-05T01:49:04.584588+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2007.14430","created_at":"2026-07-05T01:49:04.584588+00:00"},{"alias_kind":"pith_short_12","alias_value":"XPJT74A556ZM","created_at":"2026-07-05T01:49:04.584588+00:00"},{"alias_kind":"pith_short_16","alias_value":"XPJT74A556ZMM7EH","created_at":"2026-07-05T01:49:04.584588+00:00"},{"alias_kind":"pith_short_8","alias_value":"XPJT74A5","created_at":"2026-07-05T01:49:04.584588+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.01223","citing_title":"Perspectives on Tsallis Statistics for Artificial Intelligence","ref_index":32,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XPJT74A556ZMM7EH2UYWEIAYX4","json":"https://pith.science/pith/XPJT74A556ZMM7EH2UYWEIAYX4.json","graph_json":"https://pith.science/api/pith-number/XPJT74A556ZMM7EH2UYWEIAYX4/graph.json","events_json":"https://pith.science/api/pith-number/XPJT74A556ZMM7EH2UYWEIAYX4/events.json","paper":"https://pith.science/paper/XPJT74A5"},"agent_actions":{"view_html":"https://pith.science/pith/XPJT74A556ZMM7EH2UYWEIAYX4","download_json":"https://pith.science/pith/XPJT74A556ZMM7EH2UYWEIAYX4.json","view_paper":"https://pith.science/paper/XPJT74A5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2007.14430&json=true","fetch_graph":"https://pith.science/api/pith-number/XPJT74A556ZMM7EH2UYWEIAYX4/graph.json","fetch_events":"https://pith.science/api/pith-number/XPJT74A556ZMM7EH2UYWEIAYX4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XPJT74A556ZMM7EH2UYWEIAYX4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XPJT74A556ZMM7EH2UYWEIAYX4/action/storage_attestation","attest_author":"https://pith.science/pith/XPJT74A556ZMM7EH2UYWEIAYX4/action/author_attestation","sign_citation":"https://pith.science/pith/XPJT74A556ZMM7EH2UYWEIAYX4/action/citation_signature","submit_replication":"https://pith.science/pith/XPJT74A556ZMM7EH2UYWEIAYX4/action/replication_record"}},"created_at":"2026-07-05T01:49:04.584588+00:00","updated_at":"2026-07-05T01:49:04.584588+00:00"}