{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:FZ4UK7JIPKHBKXDI2FRKWIC27G","short_pith_number":"pith:FZ4UK7JI","schema_version":"1.0","canonical_sha256":"2e79457d287a8e155c68d162ab205af9a456f84fa381fc0e2aee47b56364f705","source":{"kind":"arxiv","id":"2303.05479","version":4},"attestation_state":"computed","paper":{"title":"Cal-QL: Calibrated Offline RL Pre-Training for Efficient Online Fine-Tuning","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Anikait Singh, Aviral Kumar, Chelsea Finn, Max Sobol Mark, Mitsuhiko Nakamoto, Sergey Levine, Yi Ma, Yuexiang Zhai","submitted_at":"2023-03-09T18:31:13Z","abstract_excerpt":"A compelling use case of offline reinforcement learning (RL) is to obtain a policy initialization from existing datasets followed by fast online fine-tuning with limited interaction. However, existing offline RL methods tend to behave poorly during fine-tuning. In this paper, we devise an approach for learning an effective initialization from offline data that also enables fast online fine-tuning capabilities. Our approach, calibrated Q-learning (Cal-QL), accomplishes this by learning a conservative value function initialization that underestimates the value of the learned policy from offline "},"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":"2303.05479","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.LG","submitted_at":"2023-03-09T18:31:13Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"171e3c0cfe9d45d923fe247e2c060e19125591a9844afdb1ef5bd5700172e8d0","abstract_canon_sha256":"7567bce6d39187c3c388f07ee4c971c4cd2011efa4dd78ca82786574827ccf35"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:35:36.351653Z","signature_b64":"NuCTIr6e8I7byiLfrB6ULw1whe4DxZD+GjGiquw3RaPM9M+sMGVtccOAAflYMF60icjrlLYvk4GgfFDOYKEHAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2e79457d287a8e155c68d162ab205af9a456f84fa381fc0e2aee47b56364f705","last_reissued_at":"2026-07-05T07:35:36.351173Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:35:36.351173Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cal-QL: Calibrated Offline RL Pre-Training for Efficient Online Fine-Tuning","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Anikait Singh, Aviral Kumar, Chelsea Finn, Max Sobol Mark, Mitsuhiko Nakamoto, Sergey Levine, Yi Ma, Yuexiang Zhai","submitted_at":"2023-03-09T18:31:13Z","abstract_excerpt":"A compelling use case of offline reinforcement learning (RL) is to obtain a policy initialization from existing datasets followed by fast online fine-tuning with limited interaction. However, existing offline RL methods tend to behave poorly during fine-tuning. In this paper, we devise an approach for learning an effective initialization from offline data that also enables fast online fine-tuning capabilities. Our approach, calibrated Q-learning (Cal-QL), accomplishes this by learning a conservative value function initialization that underestimates the value of the learned policy from offline "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2303.05479","kind":"arxiv","version":4},"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/2303.05479/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":"2303.05479","created_at":"2026-07-05T07:35:36.351231+00:00"},{"alias_kind":"arxiv_version","alias_value":"2303.05479v4","created_at":"2026-07-05T07:35:36.351231+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2303.05479","created_at":"2026-07-05T07:35:36.351231+00:00"},{"alias_kind":"pith_short_12","alias_value":"FZ4UK7JIPKHB","created_at":"2026-07-05T07:35:36.351231+00:00"},{"alias_kind":"pith_short_16","alias_value":"FZ4UK7JIPKHBKXDI","created_at":"2026-07-05T07:35:36.351231+00:00"},{"alias_kind":"pith_short_8","alias_value":"FZ4UK7JI","created_at":"2026-07-05T07:35:36.351231+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2507.07986","citing_title":"EXPO: Stable Reinforcement Learning with Expressive Policies","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2304.10573","citing_title":"IDQL: Implicit Q-Learning as an Actor-Critic Method with Diffusion Policies","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03065","citing_title":"OGPO: Sample Efficient Full-Finetuning of Generative Control Policies","ref_index":78,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FZ4UK7JIPKHBKXDI2FRKWIC27G","json":"https://pith.science/pith/FZ4UK7JIPKHBKXDI2FRKWIC27G.json","graph_json":"https://pith.science/api/pith-number/FZ4UK7JIPKHBKXDI2FRKWIC27G/graph.json","events_json":"https://pith.science/api/pith-number/FZ4UK7JIPKHBKXDI2FRKWIC27G/events.json","paper":"https://pith.science/paper/FZ4UK7JI"},"agent_actions":{"view_html":"https://pith.science/pith/FZ4UK7JIPKHBKXDI2FRKWIC27G","download_json":"https://pith.science/pith/FZ4UK7JIPKHBKXDI2FRKWIC27G.json","view_paper":"https://pith.science/paper/FZ4UK7JI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2303.05479&json=true","fetch_graph":"https://pith.science/api/pith-number/FZ4UK7JIPKHBKXDI2FRKWIC27G/graph.json","fetch_events":"https://pith.science/api/pith-number/FZ4UK7JIPKHBKXDI2FRKWIC27G/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FZ4UK7JIPKHBKXDI2FRKWIC27G/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FZ4UK7JIPKHBKXDI2FRKWIC27G/action/storage_attestation","attest_author":"https://pith.science/pith/FZ4UK7JIPKHBKXDI2FRKWIC27G/action/author_attestation","sign_citation":"https://pith.science/pith/FZ4UK7JIPKHBKXDI2FRKWIC27G/action/citation_signature","submit_replication":"https://pith.science/pith/FZ4UK7JIPKHBKXDI2FRKWIC27G/action/replication_record"}},"created_at":"2026-07-05T07:35:36.351231+00:00","updated_at":"2026-07-05T07:35:36.351231+00:00"}