{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:IF3MD7QO7VSKKZ6SIF3KF4VEFW","short_pith_number":"pith:IF3MD7QO","schema_version":"1.0","canonical_sha256":"4176c1fe0efd64a567d24176a2f2a42dbda4456213c62ad6db68dd78519de5b5","source":{"kind":"arxiv","id":"2402.00348","version":1},"attestation_state":"computed","paper":{"title":"ODICE: Revealing the Mystery of Distribution Correction Estimation via Orthogonal-gradient Update","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Haoran Xu, Liyuan Mao, Weinan Zhang, Xianyuan Zhan","submitted_at":"2024-02-01T05:30:51Z","abstract_excerpt":"In this study, we investigate the DIstribution Correction Estimation (DICE) methods, an important line of work in offline reinforcement learning (RL) and imitation learning (IL). DICE-based methods impose state-action-level behavior constraint, which is an ideal choice for offline learning. However, they typically perform much worse than current state-of-the-art (SOTA) methods that solely use action-level behavior constraint. After revisiting DICE-based methods, we find there exist two gradient terms when learning the value function using true-gradient update: forward gradient (taken on the cu"},"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":"2402.00348","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-01T05:30:51Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f47371489d7a5495c38b016cbe5be900fd844636df739f1bab758eccc4cce1d7","abstract_canon_sha256":"16e08c1547d976d4a0ceb136737e97759a1aa78dbd97c4f578fc77cc83410976"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:40:10.416542Z","signature_b64":"0EjelexYbVbjFOakmANvyR8Tpjq0ch9i3i78TFaOxCCz7w+mA5fwXkslx9v/RH6L3no+PJBQ6zOXnijTVdsfCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4176c1fe0efd64a567d24176a2f2a42dbda4456213c62ad6db68dd78519de5b5","last_reissued_at":"2026-07-05T07:40:10.416048Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:40:10.416048Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ODICE: Revealing the Mystery of Distribution Correction Estimation via Orthogonal-gradient Update","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Haoran Xu, Liyuan Mao, Weinan Zhang, Xianyuan Zhan","submitted_at":"2024-02-01T05:30:51Z","abstract_excerpt":"In this study, we investigate the DIstribution Correction Estimation (DICE) methods, an important line of work in offline reinforcement learning (RL) and imitation learning (IL). DICE-based methods impose state-action-level behavior constraint, which is an ideal choice for offline learning. However, they typically perform much worse than current state-of-the-art (SOTA) methods that solely use action-level behavior constraint. After revisiting DICE-based methods, we find there exist two gradient terms when learning the value function using true-gradient update: forward gradient (taken on the cu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.00348","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":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2402.00348/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":"2402.00348","created_at":"2026-07-05T07:40:10.416117+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.00348v1","created_at":"2026-07-05T07:40:10.416117+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.00348","created_at":"2026-07-05T07:40:10.416117+00:00"},{"alias_kind":"pith_short_12","alias_value":"IF3MD7QO7VSK","created_at":"2026-07-05T07:40:10.416117+00:00"},{"alias_kind":"pith_short_16","alias_value":"IF3MD7QO7VSKKZ6S","created_at":"2026-07-05T07:40:10.416117+00:00"},{"alias_kind":"pith_short_8","alias_value":"IF3MD7QO","created_at":"2026-07-05T07:40:10.416117+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2504.13368","citing_title":"An Optimal Discriminator Weighted Imitation Perspective for Reinforcement Learning","ref_index":1960,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IF3MD7QO7VSKKZ6SIF3KF4VEFW","json":"https://pith.science/pith/IF3MD7QO7VSKKZ6SIF3KF4VEFW.json","graph_json":"https://pith.science/api/pith-number/IF3MD7QO7VSKKZ6SIF3KF4VEFW/graph.json","events_json":"https://pith.science/api/pith-number/IF3MD7QO7VSKKZ6SIF3KF4VEFW/events.json","paper":"https://pith.science/paper/IF3MD7QO"},"agent_actions":{"view_html":"https://pith.science/pith/IF3MD7QO7VSKKZ6SIF3KF4VEFW","download_json":"https://pith.science/pith/IF3MD7QO7VSKKZ6SIF3KF4VEFW.json","view_paper":"https://pith.science/paper/IF3MD7QO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.00348&json=true","fetch_graph":"https://pith.science/api/pith-number/IF3MD7QO7VSKKZ6SIF3KF4VEFW/graph.json","fetch_events":"https://pith.science/api/pith-number/IF3MD7QO7VSKKZ6SIF3KF4VEFW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IF3MD7QO7VSKKZ6SIF3KF4VEFW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IF3MD7QO7VSKKZ6SIF3KF4VEFW/action/storage_attestation","attest_author":"https://pith.science/pith/IF3MD7QO7VSKKZ6SIF3KF4VEFW/action/author_attestation","sign_citation":"https://pith.science/pith/IF3MD7QO7VSKKZ6SIF3KF4VEFW/action/citation_signature","submit_replication":"https://pith.science/pith/IF3MD7QO7VSKKZ6SIF3KF4VEFW/action/replication_record"}},"created_at":"2026-07-05T07:40:10.416117+00:00","updated_at":"2026-07-05T07:40:10.416117+00:00"}