{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:OJADGW2DZIISOXDC7HGT4GN22G","short_pith_number":"pith:OJADGW2D","schema_version":"1.0","canonical_sha256":"7240335b43ca11275c62f9cd3e19bad18fd42a674cadb1c3ef25be656274a2e7","source":{"kind":"arxiv","id":"2410.09361","version":1},"attestation_state":"computed","paper":{"title":"Decision-Point Guided Safe Policy Improvement","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Abhishek Sharma, Finale Doshi-Velez, Leo Benac, Sonali Parbhoo","submitted_at":"2024-10-12T04:05:56Z","abstract_excerpt":"Within batch reinforcement learning, safe policy improvement (SPI) seeks to ensure that the learnt policy performs at least as well as the behavior policy that generated the dataset. The core challenge in SPI is seeking improvements while balancing risk when many state-action pairs may be infrequently visited. In this work, we introduce Decision Points RL (DPRL), an algorithm that restricts the set of state-action pairs (or regions for continuous states) considered for improvement. DPRL ensures high-confidence improvement in densely visited states (i.e. decision points) while still utilizing d"},"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":"2410.09361","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-12T04:05:56Z","cross_cats_sorted":[],"title_canon_sha256":"a731f99c054a1586f8bd2d3eba9990c4f5349584323434b8b6a3c878d125440c","abstract_canon_sha256":"dc373a1ab2bde4f565725fa398a1526af1aea4cab5b068da6078b76078840922"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:19:38.188968Z","signature_b64":"0apPc8EGSku89Is3UHlwfe/KFZtdTUVdDSB1H56ITPt+QEwC0vU/C6AGCSVfFh5ncin1uGKXUtP+KpkYuD8cAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7240335b43ca11275c62f9cd3e19bad18fd42a674cadb1c3ef25be656274a2e7","last_reissued_at":"2026-07-05T09:19:38.188576Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:19:38.188576Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Decision-Point Guided Safe Policy Improvement","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Abhishek Sharma, Finale Doshi-Velez, Leo Benac, Sonali Parbhoo","submitted_at":"2024-10-12T04:05:56Z","abstract_excerpt":"Within batch reinforcement learning, safe policy improvement (SPI) seeks to ensure that the learnt policy performs at least as well as the behavior policy that generated the dataset. The core challenge in SPI is seeking improvements while balancing risk when many state-action pairs may be infrequently visited. In this work, we introduce Decision Points RL (DPRL), an algorithm that restricts the set of state-action pairs (or regions for continuous states) considered for improvement. DPRL ensures high-confidence improvement in densely visited states (i.e. decision points) while still utilizing d"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.09361","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/2410.09361/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":"2410.09361","created_at":"2026-07-05T09:19:38.188635+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.09361v1","created_at":"2026-07-05T09:19:38.188635+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.09361","created_at":"2026-07-05T09:19:38.188635+00:00"},{"alias_kind":"pith_short_12","alias_value":"OJADGW2DZIIS","created_at":"2026-07-05T09:19:38.188635+00:00"},{"alias_kind":"pith_short_16","alias_value":"OJADGW2DZIISOXDC","created_at":"2026-07-05T09:19:38.188635+00:00"},{"alias_kind":"pith_short_8","alias_value":"OJADGW2D","created_at":"2026-07-05T09:19:38.188635+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.16833","citing_title":"Strategically Linked Decisions in Long-Term Planning and Reinforcement Learning","ref_index":20,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OJADGW2DZIISOXDC7HGT4GN22G","json":"https://pith.science/pith/OJADGW2DZIISOXDC7HGT4GN22G.json","graph_json":"https://pith.science/api/pith-number/OJADGW2DZIISOXDC7HGT4GN22G/graph.json","events_json":"https://pith.science/api/pith-number/OJADGW2DZIISOXDC7HGT4GN22G/events.json","paper":"https://pith.science/paper/OJADGW2D"},"agent_actions":{"view_html":"https://pith.science/pith/OJADGW2DZIISOXDC7HGT4GN22G","download_json":"https://pith.science/pith/OJADGW2DZIISOXDC7HGT4GN22G.json","view_paper":"https://pith.science/paper/OJADGW2D","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.09361&json=true","fetch_graph":"https://pith.science/api/pith-number/OJADGW2DZIISOXDC7HGT4GN22G/graph.json","fetch_events":"https://pith.science/api/pith-number/OJADGW2DZIISOXDC7HGT4GN22G/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OJADGW2DZIISOXDC7HGT4GN22G/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OJADGW2DZIISOXDC7HGT4GN22G/action/storage_attestation","attest_author":"https://pith.science/pith/OJADGW2DZIISOXDC7HGT4GN22G/action/author_attestation","sign_citation":"https://pith.science/pith/OJADGW2DZIISOXDC7HGT4GN22G/action/citation_signature","submit_replication":"https://pith.science/pith/OJADGW2DZIISOXDC7HGT4GN22G/action/replication_record"}},"created_at":"2026-07-05T09:19:38.188635+00:00","updated_at":"2026-07-05T09:19:38.188635+00:00"}