{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2017:ZYINZIHLQNLUMIPSSYSJWFXT62","short_pith_number":"pith:ZYINZIHL","schema_version":"1.0","canonical_sha256":"ce10dca0eb83574621f296249b16f3f6909ce215977735484e2f6d2a8f1e79f1","source":{"kind":"arxiv","id":"1801.00056","version":2},"attestation_state":"computed","paper":{"title":"f-Divergence constrained policy improvement","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Boris Belousov, Jan Peters","submitted_at":"2017-12-29T23:07:26Z","abstract_excerpt":"To ensure stability of learning, state-of-the-art generalized policy iteration algorithms augment the policy improvement step with a trust region constraint bounding the information loss. The size of the trust region is commonly determined by the Kullback-Leibler (KL) divergence, which not only captures the notion of distance well but also yields closed-form solutions. In this paper, we consider a more general class of f-divergences and derive the corresponding policy update rules. The generic solution is expressed through the derivative of the convex conjugate function to f and includes the K"},"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":"1801.00056","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2017-12-29T23:07:26Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"8079f8e09b7c215f6e6d142a69f9504f3b53b80bedebd95dad0f6f3a13819a24","abstract_canon_sha256":"35bb8de6b9eda3543155d9e40c8c602635a195a518394a89dcd33c743308cdb1"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:19:16.592775Z","signature_b64":"5sgP54RC775zoQgdMt4kKbSVIfmnCLxduuKv/X1icEL6fl+tOR6VD+3SlotQa5pJ9/oi9bYLmA2Lh4TEm7JzBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ce10dca0eb83574621f296249b16f3f6909ce215977735484e2f6d2a8f1e79f1","last_reissued_at":"2026-05-18T00:19:16.592171Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:19:16.592171Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"f-Divergence constrained policy improvement","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Boris Belousov, Jan Peters","submitted_at":"2017-12-29T23:07:26Z","abstract_excerpt":"To ensure stability of learning, state-of-the-art generalized policy iteration algorithms augment the policy improvement step with a trust region constraint bounding the information loss. The size of the trust region is commonly determined by the Kullback-Leibler (KL) divergence, which not only captures the notion of distance well but also yields closed-form solutions. In this paper, we consider a more general class of f-divergences and derive the corresponding policy update rules. The generic solution is expressed through the derivative of the convex conjugate function to f and includes the K"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1801.00056","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":""},"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":"1801.00056","created_at":"2026-05-18T00:19:16.592246+00:00"},{"alias_kind":"arxiv_version","alias_value":"1801.00056v2","created_at":"2026-05-18T00:19:16.592246+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1801.00056","created_at":"2026-05-18T00:19:16.592246+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZYINZIHLQNLU","created_at":"2026-05-18T12:31:59.375834+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZYINZIHLQNLUMIPS","created_at":"2026-05-18T12:31:59.375834+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZYINZIHL","created_at":"2026-05-18T12:31:59.375834+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.06977","citing_title":"$f$-Divergence Regularized RLHF: Two Tales of Sampling and Unified Analyses","ref_index":3,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZYINZIHLQNLUMIPSSYSJWFXT62","json":"https://pith.science/pith/ZYINZIHLQNLUMIPSSYSJWFXT62.json","graph_json":"https://pith.science/api/pith-number/ZYINZIHLQNLUMIPSSYSJWFXT62/graph.json","events_json":"https://pith.science/api/pith-number/ZYINZIHLQNLUMIPSSYSJWFXT62/events.json","paper":"https://pith.science/paper/ZYINZIHL"},"agent_actions":{"view_html":"https://pith.science/pith/ZYINZIHLQNLUMIPSSYSJWFXT62","download_json":"https://pith.science/pith/ZYINZIHLQNLUMIPSSYSJWFXT62.json","view_paper":"https://pith.science/paper/ZYINZIHL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1801.00056&json=true","fetch_graph":"https://pith.science/api/pith-number/ZYINZIHLQNLUMIPSSYSJWFXT62/graph.json","fetch_events":"https://pith.science/api/pith-number/ZYINZIHLQNLUMIPSSYSJWFXT62/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZYINZIHLQNLUMIPSSYSJWFXT62/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZYINZIHLQNLUMIPSSYSJWFXT62/action/storage_attestation","attest_author":"https://pith.science/pith/ZYINZIHLQNLUMIPSSYSJWFXT62/action/author_attestation","sign_citation":"https://pith.science/pith/ZYINZIHLQNLUMIPSSYSJWFXT62/action/citation_signature","submit_replication":"https://pith.science/pith/ZYINZIHLQNLUMIPSSYSJWFXT62/action/replication_record"}},"created_at":"2026-05-18T00:19:16.592246+00:00","updated_at":"2026-05-18T00:19:16.592246+00:00"}