{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:JBP2HCEXSXRQYMAH6FM74PWF3U","short_pith_number":"pith:JBP2HCEX","schema_version":"1.0","canonical_sha256":"485fa3889795e30c3007f159fe3ec5dd0ff0d332a4bb534ec9b2d373a8497aea","source":{"kind":"arxiv","id":"2104.09469","version":1},"attestation_state":"computed","paper":{"title":"Training Value-Aligned Reinforcement Learning Agents Using a Normative Prior","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.HC"],"primary_cat":"cs.LG","authors_text":"Brent Harrison, Mark Riedl, Md Sultan Al Nahian, Spencer Frazier","submitted_at":"2021-04-19T17:33:07Z","abstract_excerpt":"As more machine learning agents interact with humans, it is increasingly a prospect that an agent trained to perform a task optimally, using only a measure of task performance as feedback, can violate societal norms for acceptable behavior or cause harm. Value alignment is a property of intelligent agents wherein they solely pursue non-harmful behaviors or human-beneficial goals. We introduce an approach to value-aligned reinforcement learning, in which we train an agent with two reward signals: a standard task performance reward, plus a normative behavior reward. The normative behavior reward"},"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":"2104.09469","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2021-04-19T17:33:07Z","cross_cats_sorted":["cs.AI","cs.HC"],"title_canon_sha256":"616a32c5adcfb01e2b5cb64fb8e3bf73ccbe4034e27e6840212fd82d7149832c","abstract_canon_sha256":"979c2453543142ad562c2d9413fd1a585c290d69f3245a864e496b16afff0916"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:33:13.592462Z","signature_b64":"DBs9tk803iA0Fk+DbCpq7aAVIWowkTpmoMBFOHq3GIQuiu7TIKyDZYlBAC6k79LDNVaWd0afmH0mVpMiGJBCBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"485fa3889795e30c3007f159fe3ec5dd0ff0d332a4bb534ec9b2d373a8497aea","last_reissued_at":"2026-07-05T02:33:13.591986Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:33:13.591986Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Training Value-Aligned Reinforcement Learning Agents Using a Normative Prior","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.HC"],"primary_cat":"cs.LG","authors_text":"Brent Harrison, Mark Riedl, Md Sultan Al Nahian, Spencer Frazier","submitted_at":"2021-04-19T17:33:07Z","abstract_excerpt":"As more machine learning agents interact with humans, it is increasingly a prospect that an agent trained to perform a task optimally, using only a measure of task performance as feedback, can violate societal norms for acceptable behavior or cause harm. Value alignment is a property of intelligent agents wherein they solely pursue non-harmful behaviors or human-beneficial goals. We introduce an approach to value-aligned reinforcement learning, in which we train an agent with two reward signals: a standard task performance reward, plus a normative behavior reward. The normative behavior reward"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2104.09469","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/2104.09469/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":"2104.09469","created_at":"2026-07-05T02:33:13.592043+00:00"},{"alias_kind":"arxiv_version","alias_value":"2104.09469v1","created_at":"2026-07-05T02:33:13.592043+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2104.09469","created_at":"2026-07-05T02:33:13.592043+00:00"},{"alias_kind":"pith_short_12","alias_value":"JBP2HCEXSXRQ","created_at":"2026-07-05T02:33:13.592043+00:00"},{"alias_kind":"pith_short_16","alias_value":"JBP2HCEXSXRQYMAH","created_at":"2026-07-05T02:33:13.592043+00:00"},{"alias_kind":"pith_short_8","alias_value":"JBP2HCEX","created_at":"2026-07-05T02:33:13.592043+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.15011","citing_title":"HAVA: Hybrid Approach to Value-Alignment through Reward Weighing for Reinforcement Learning","ref_index":17,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JBP2HCEXSXRQYMAH6FM74PWF3U","json":"https://pith.science/pith/JBP2HCEXSXRQYMAH6FM74PWF3U.json","graph_json":"https://pith.science/api/pith-number/JBP2HCEXSXRQYMAH6FM74PWF3U/graph.json","events_json":"https://pith.science/api/pith-number/JBP2HCEXSXRQYMAH6FM74PWF3U/events.json","paper":"https://pith.science/paper/JBP2HCEX"},"agent_actions":{"view_html":"https://pith.science/pith/JBP2HCEXSXRQYMAH6FM74PWF3U","download_json":"https://pith.science/pith/JBP2HCEXSXRQYMAH6FM74PWF3U.json","view_paper":"https://pith.science/paper/JBP2HCEX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2104.09469&json=true","fetch_graph":"https://pith.science/api/pith-number/JBP2HCEXSXRQYMAH6FM74PWF3U/graph.json","fetch_events":"https://pith.science/api/pith-number/JBP2HCEXSXRQYMAH6FM74PWF3U/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JBP2HCEXSXRQYMAH6FM74PWF3U/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JBP2HCEXSXRQYMAH6FM74PWF3U/action/storage_attestation","attest_author":"https://pith.science/pith/JBP2HCEXSXRQYMAH6FM74PWF3U/action/author_attestation","sign_citation":"https://pith.science/pith/JBP2HCEXSXRQYMAH6FM74PWF3U/action/citation_signature","submit_replication":"https://pith.science/pith/JBP2HCEXSXRQYMAH6FM74PWF3U/action/replication_record"}},"created_at":"2026-07-05T02:33:13.592043+00:00","updated_at":"2026-07-05T02:33:13.592043+00:00"}