{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:LNUPY4HRBGRTCOW6PRRXVYZQ5D","short_pith_number":"pith:LNUPY4HR","schema_version":"1.0","canonical_sha256":"5b68fc70f109a3313ade7c637ae330e8d1773893715c7f3cdc228aee051e7708","source":{"kind":"arxiv","id":"2102.05371","version":1},"attestation_state":"computed","paper":{"title":"Risk-Averse Offline Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andreas Krause, N\\'uria Armengol Urp\\'i, Sebastian Curi","submitted_at":"2021-02-10T10:27:49Z","abstract_excerpt":"Training Reinforcement Learning (RL) agents in high-stakes applications might be too prohibitive due to the risk associated to exploration. Thus, the agent can only use data previously collected by safe policies. While previous work considers optimizing the average performance using offline data, we focus on optimizing a risk-averse criteria, namely the CVaR. In particular, we present the Offline Risk-Averse Actor-Critic (O-RAAC), a model-free RL algorithm that is able to learn risk-averse policies in a fully offline setting. We show that O-RAAC learns policies with higher CVaR than risk-neutr"},"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":"2102.05371","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2021-02-10T10:27:49Z","cross_cats_sorted":[],"title_canon_sha256":"92d90fb23e8f9ffec20da8bd06ce92401c1cee34a6ad82f778080ed5720c851f","abstract_canon_sha256":"24c9110a3d15bac6969b7883386bd1a72966d3c0044be8578d9e2aefe06568a2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:14:17.984993Z","signature_b64":"vTogqxqnl4/O5Vautk08oypt6cVBE7Z3Uu9+kJZeTnWlnpIFVO9vb4EibSdeAut4O7YAKWD7a9+v6pWYhXe7Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5b68fc70f109a3313ade7c637ae330e8d1773893715c7f3cdc228aee051e7708","last_reissued_at":"2026-07-05T02:14:17.984558Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:14:17.984558Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Risk-Averse Offline Reinforcement Learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andreas Krause, N\\'uria Armengol Urp\\'i, Sebastian Curi","submitted_at":"2021-02-10T10:27:49Z","abstract_excerpt":"Training Reinforcement Learning (RL) agents in high-stakes applications might be too prohibitive due to the risk associated to exploration. Thus, the agent can only use data previously collected by safe policies. While previous work considers optimizing the average performance using offline data, we focus on optimizing a risk-averse criteria, namely the CVaR. In particular, we present the Offline Risk-Averse Actor-Critic (O-RAAC), a model-free RL algorithm that is able to learn risk-averse policies in a fully offline setting. We show that O-RAAC learns policies with higher CVaR than risk-neutr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2102.05371","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/2102.05371/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":"2102.05371","created_at":"2026-07-05T02:14:17.984629+00:00"},{"alias_kind":"arxiv_version","alias_value":"2102.05371v1","created_at":"2026-07-05T02:14:17.984629+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2102.05371","created_at":"2026-07-05T02:14:17.984629+00:00"},{"alias_kind":"pith_short_12","alias_value":"LNUPY4HRBGRT","created_at":"2026-07-05T02:14:17.984629+00:00"},{"alias_kind":"pith_short_16","alias_value":"LNUPY4HRBGRTCOW6","created_at":"2026-07-05T02:14:17.984629+00:00"},{"alias_kind":"pith_short_8","alias_value":"LNUPY4HR","created_at":"2026-07-05T02:14:17.984629+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11249","citing_title":"MASK: Multi-Agent Semantic K-Scheduling for Risk-Sensitive 6G Robotics","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2606.27766","citing_title":"RS-Diffuser: Risk-Sensitive Diffusion Planning with Distributional Value Guidance","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01663","citing_title":"Towards Efficient and Expressive Offline RL via Flow-Anchored Noise-conditioned Q-Learning","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LNUPY4HRBGRTCOW6PRRXVYZQ5D","json":"https://pith.science/pith/LNUPY4HRBGRTCOW6PRRXVYZQ5D.json","graph_json":"https://pith.science/api/pith-number/LNUPY4HRBGRTCOW6PRRXVYZQ5D/graph.json","events_json":"https://pith.science/api/pith-number/LNUPY4HRBGRTCOW6PRRXVYZQ5D/events.json","paper":"https://pith.science/paper/LNUPY4HR"},"agent_actions":{"view_html":"https://pith.science/pith/LNUPY4HRBGRTCOW6PRRXVYZQ5D","download_json":"https://pith.science/pith/LNUPY4HRBGRTCOW6PRRXVYZQ5D.json","view_paper":"https://pith.science/paper/LNUPY4HR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2102.05371&json=true","fetch_graph":"https://pith.science/api/pith-number/LNUPY4HRBGRTCOW6PRRXVYZQ5D/graph.json","fetch_events":"https://pith.science/api/pith-number/LNUPY4HRBGRTCOW6PRRXVYZQ5D/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LNUPY4HRBGRTCOW6PRRXVYZQ5D/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LNUPY4HRBGRTCOW6PRRXVYZQ5D/action/storage_attestation","attest_author":"https://pith.science/pith/LNUPY4HRBGRTCOW6PRRXVYZQ5D/action/author_attestation","sign_citation":"https://pith.science/pith/LNUPY4HRBGRTCOW6PRRXVYZQ5D/action/citation_signature","submit_replication":"https://pith.science/pith/LNUPY4HRBGRTCOW6PRRXVYZQ5D/action/replication_record"}},"created_at":"2026-07-05T02:14:17.984629+00:00","updated_at":"2026-07-05T02:14:17.984629+00:00"}