{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:DCRWFIXSLVS2JSW66UTRVKRXUX","short_pith_number":"pith:DCRWFIXS","schema_version":"1.0","canonical_sha256":"18a362a2f25d65a4cadef5271aaa37a5e47bb1595ec1b1d6a9db9998ac18f73a","source":{"kind":"arxiv","id":"2003.14089","version":5},"attestation_state":"computed","paper":{"title":"Leverage the Average: an Analysis of KL Regularization in RL","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Bruno Scherrer, Matthieu Geist, Nino Vieillard, Olivier Pietquin, R\\'emi Munos, Tadashi Kozuno","submitted_at":"2020-03-31T10:55:06Z","abstract_excerpt":"Recent Reinforcement Learning (RL) algorithms making use of Kullback-Leibler (KL) regularization as a core component have shown outstanding performance. Yet, only little is understood theoretically about why KL regularization helps, so far. We study KL regularization within an approximate value iteration scheme and show that it implicitly averages q-values. Leveraging this insight, we provide a very strong performance bound, the very first to combine two desirable aspects: a linear dependency to the horizon (instead of quadratic) and an error propagation term involving an averaging effect of t"},"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":"2003.14089","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2020-03-31T10:55:06Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"e00d129258d193c1596dd17002397eaf16bed93ee2158554154a6712ada0b1c2","abstract_canon_sha256":"c2956d9ce76693b10a42d9c82714b7f4af11bf846356e6cdc828041db01d6881"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:04:56.168973Z","signature_b64":"VfQVyvsdkq6llsbWEO5Trr12YrnNmkRmaeu1W3smShA5kOGFeI61mxUL1/H0Z8kWkh9o3tFjc96jNFfm4yInAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"18a362a2f25d65a4cadef5271aaa37a5e47bb1595ec1b1d6a9db9998ac18f73a","last_reissued_at":"2026-07-05T02:04:56.168429Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:04:56.168429Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Leverage the Average: an Analysis of KL Regularization in RL","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Bruno Scherrer, Matthieu Geist, Nino Vieillard, Olivier Pietquin, R\\'emi Munos, Tadashi Kozuno","submitted_at":"2020-03-31T10:55:06Z","abstract_excerpt":"Recent Reinforcement Learning (RL) algorithms making use of Kullback-Leibler (KL) regularization as a core component have shown outstanding performance. Yet, only little is understood theoretically about why KL regularization helps, so far. We study KL regularization within an approximate value iteration scheme and show that it implicitly averages q-values. Leveraging this insight, we provide a very strong performance bound, the very first to combine two desirable aspects: a linear dependency to the horizon (instead of quadratic) and an error propagation term involving an averaging effect of t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2003.14089","kind":"arxiv","version":5},"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/2003.14089/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":"2003.14089","created_at":"2026-07-05T02:04:56.168488+00:00"},{"alias_kind":"arxiv_version","alias_value":"2003.14089v5","created_at":"2026-07-05T02:04:56.168488+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2003.14089","created_at":"2026-07-05T02:04:56.168488+00:00"},{"alias_kind":"pith_short_12","alias_value":"DCRWFIXSLVS2","created_at":"2026-07-05T02:04:56.168488+00:00"},{"alias_kind":"pith_short_16","alias_value":"DCRWFIXSLVS2JSW6","created_at":"2026-07-05T02:04:56.168488+00:00"},{"alias_kind":"pith_short_8","alias_value":"DCRWFIXS","created_at":"2026-07-05T02:04:56.168488+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DCRWFIXSLVS2JSW66UTRVKRXUX","json":"https://pith.science/pith/DCRWFIXSLVS2JSW66UTRVKRXUX.json","graph_json":"https://pith.science/api/pith-number/DCRWFIXSLVS2JSW66UTRVKRXUX/graph.json","events_json":"https://pith.science/api/pith-number/DCRWFIXSLVS2JSW66UTRVKRXUX/events.json","paper":"https://pith.science/paper/DCRWFIXS"},"agent_actions":{"view_html":"https://pith.science/pith/DCRWFIXSLVS2JSW66UTRVKRXUX","download_json":"https://pith.science/pith/DCRWFIXSLVS2JSW66UTRVKRXUX.json","view_paper":"https://pith.science/paper/DCRWFIXS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2003.14089&json=true","fetch_graph":"https://pith.science/api/pith-number/DCRWFIXSLVS2JSW66UTRVKRXUX/graph.json","fetch_events":"https://pith.science/api/pith-number/DCRWFIXSLVS2JSW66UTRVKRXUX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DCRWFIXSLVS2JSW66UTRVKRXUX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DCRWFIXSLVS2JSW66UTRVKRXUX/action/storage_attestation","attest_author":"https://pith.science/pith/DCRWFIXSLVS2JSW66UTRVKRXUX/action/author_attestation","sign_citation":"https://pith.science/pith/DCRWFIXSLVS2JSW66UTRVKRXUX/action/citation_signature","submit_replication":"https://pith.science/pith/DCRWFIXSLVS2JSW66UTRVKRXUX/action/replication_record"}},"created_at":"2026-07-05T02:04:56.168488+00:00","updated_at":"2026-07-05T02:04:56.168488+00:00"}