{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:YP2GWEONAJOZAUEVR5A7EAGA55","short_pith_number":"pith:YP2GWEON","schema_version":"1.0","canonical_sha256":"c3f46b11cd025d9050958f41f200c0ef74d2a8e7f4dc3bf4dc615afb51d611be","source":{"kind":"arxiv","id":"1810.12429","version":1},"attestation_state":"computed","paper":{"title":"Breaking the Curse of Horizon: Infinite-Horizon Off-Policy Estimation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.SY","stat.ML"],"primary_cat":"cs.LG","authors_text":"Dengyong Zhou, Lihong Li, Qiang Liu, Ziyang Tang","submitted_at":"2018-10-29T22:03:58Z","abstract_excerpt":"We consider the off-policy estimation problem of estimating the expected reward of a target policy using samples collected by a different behavior policy. Importance sampling (IS) has been a key technique to derive (nearly) unbiased estimators, but is known to suffer from an excessively high variance in long-horizon problems. In the extreme case of in infinite-horizon problems, the variance of an IS-based estimator may even be unbounded. In this paper, we propose a new off-policy estimation method that applies IS directly on the stationary state-visitation distributions to avoid the exploding "},"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":"1810.12429","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2018-10-29T22:03:58Z","cross_cats_sorted":["cs.AI","cs.SY","stat.ML"],"title_canon_sha256":"b2ea5f7578410c8d6104be7c23b4635fea014eb99b71a83b56f1409091f568d4","abstract_canon_sha256":"f7e0a236bc5ea99ad44c343f6bce7139009d21b2d87a7cc823bd2f6eba8fd502"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:02:02.689134Z","signature_b64":"vwv/b2u4qMJMa81QcjP6kUime3DO0pOUoXBvK582SIT95smASOnRx0JR3QgQM5cbvyAOII09BBEQX0NRZR2qBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c3f46b11cd025d9050958f41f200c0ef74d2a8e7f4dc3bf4dc615afb51d611be","last_reissued_at":"2026-05-18T00:02:02.688608Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:02:02.688608Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Breaking the Curse of Horizon: Infinite-Horizon Off-Policy Estimation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.SY","stat.ML"],"primary_cat":"cs.LG","authors_text":"Dengyong Zhou, Lihong Li, Qiang Liu, Ziyang Tang","submitted_at":"2018-10-29T22:03:58Z","abstract_excerpt":"We consider the off-policy estimation problem of estimating the expected reward of a target policy using samples collected by a different behavior policy. Importance sampling (IS) has been a key technique to derive (nearly) unbiased estimators, but is known to suffer from an excessively high variance in long-horizon problems. In the extreme case of in infinite-horizon problems, the variance of an IS-based estimator may even be unbounded. In this paper, we propose a new off-policy estimation method that applies IS directly on the stationary state-visitation distributions to avoid the exploding "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1810.12429","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":""},"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":"1810.12429","created_at":"2026-05-18T00:02:02.688696+00:00"},{"alias_kind":"arxiv_version","alias_value":"1810.12429v1","created_at":"2026-05-18T00:02:02.688696+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1810.12429","created_at":"2026-05-18T00:02:02.688696+00:00"},{"alias_kind":"pith_short_12","alias_value":"YP2GWEONAJOZ","created_at":"2026-05-18T12:33:04.347982+00:00"},{"alias_kind":"pith_short_16","alias_value":"YP2GWEONAJOZAUEV","created_at":"2026-05-18T12:33:04.347982+00:00"},{"alias_kind":"pith_short_8","alias_value":"YP2GWEON","created_at":"2026-05-18T12:33:04.347982+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.08441","citing_title":"DUET: Optimize Token-Budget Allocation for Reinforcement Learning with Verifiable Rewards","ref_index":20,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YP2GWEONAJOZAUEVR5A7EAGA55","json":"https://pith.science/pith/YP2GWEONAJOZAUEVR5A7EAGA55.json","graph_json":"https://pith.science/api/pith-number/YP2GWEONAJOZAUEVR5A7EAGA55/graph.json","events_json":"https://pith.science/api/pith-number/YP2GWEONAJOZAUEVR5A7EAGA55/events.json","paper":"https://pith.science/paper/YP2GWEON"},"agent_actions":{"view_html":"https://pith.science/pith/YP2GWEONAJOZAUEVR5A7EAGA55","download_json":"https://pith.science/pith/YP2GWEONAJOZAUEVR5A7EAGA55.json","view_paper":"https://pith.science/paper/YP2GWEON","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1810.12429&json=true","fetch_graph":"https://pith.science/api/pith-number/YP2GWEONAJOZAUEVR5A7EAGA55/graph.json","fetch_events":"https://pith.science/api/pith-number/YP2GWEONAJOZAUEVR5A7EAGA55/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YP2GWEONAJOZAUEVR5A7EAGA55/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YP2GWEONAJOZAUEVR5A7EAGA55/action/storage_attestation","attest_author":"https://pith.science/pith/YP2GWEONAJOZAUEVR5A7EAGA55/action/author_attestation","sign_citation":"https://pith.science/pith/YP2GWEONAJOZAUEVR5A7EAGA55/action/citation_signature","submit_replication":"https://pith.science/pith/YP2GWEONAJOZAUEVR5A7EAGA55/action/replication_record"}},"created_at":"2026-05-18T00:02:02.688696+00:00","updated_at":"2026-05-18T00:02:02.688696+00:00"}