{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:OHUJH2YVDRMMNTHZGNQ4K62FOE","short_pith_number":"pith:OHUJH2YV","schema_version":"1.0","canonical_sha256":"71e893eb151c58c6ccf93361c57b457135a9e83b7e0d125f2db71c7ce6bed72b","source":{"kind":"arxiv","id":"2207.06269","version":1},"attestation_state":"computed","paper":{"title":"Policy Optimization with Sparse Global Contrastive Explanations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Finale Doshi-Velez, Jiayu Yao, Sonali Parbhoo, Weiwei Pan","submitted_at":"2022-07-13T15:17:26Z","abstract_excerpt":"We develop a Reinforcement Learning (RL) framework for improving an existing behavior policy via sparse, user-interpretable changes. Our goal is to make minimal changes while gaining as much benefit as possible. We define a minimal change as having a sparse, global contrastive explanation between the original and proposed policy. We improve the current policy with the constraint of keeping that global contrastive explanation short. We demonstrate our framework with a discrete MDP and a continuous 2D navigation domain."},"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":"2207.06269","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-07-13T15:17:26Z","cross_cats_sorted":[],"title_canon_sha256":"34e4951ccf58da4e0b8f7cfc8efd93616f68a77f551cb87315310edab2b372ae","abstract_canon_sha256":"ce08cf662a6a9ec7cc283c8e12049b38b50ef466d9ef53afde78fc50c87f5e22"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:40:03.388805Z","signature_b64":"4sapHZsgwU3mbz1pJU7M+2mZBijvrMnU6TnndgbQHXENKx4WIu2blH/dfONDDUeVG7jopP/K9cyByQu7FYdnCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"71e893eb151c58c6ccf93361c57b457135a9e83b7e0d125f2db71c7ce6bed72b","last_reissued_at":"2026-07-05T04:40:03.388247Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:40:03.388247Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Policy Optimization with Sparse Global Contrastive Explanations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Finale Doshi-Velez, Jiayu Yao, Sonali Parbhoo, Weiwei Pan","submitted_at":"2022-07-13T15:17:26Z","abstract_excerpt":"We develop a Reinforcement Learning (RL) framework for improving an existing behavior policy via sparse, user-interpretable changes. Our goal is to make minimal changes while gaining as much benefit as possible. We define a minimal change as having a sparse, global contrastive explanation between the original and proposed policy. We improve the current policy with the constraint of keeping that global contrastive explanation short. We demonstrate our framework with a discrete MDP and a continuous 2D navigation domain."},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2207.06269","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/2207.06269/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":"2207.06269","created_at":"2026-07-05T04:40:03.388313+00:00"},{"alias_kind":"arxiv_version","alias_value":"2207.06269v1","created_at":"2026-07-05T04:40:03.388313+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2207.06269","created_at":"2026-07-05T04:40:03.388313+00:00"},{"alias_kind":"pith_short_12","alias_value":"OHUJH2YVDRMM","created_at":"2026-07-05T04:40:03.388313+00:00"},{"alias_kind":"pith_short_16","alias_value":"OHUJH2YVDRMMNTHZ","created_at":"2026-07-05T04:40:03.388313+00:00"},{"alias_kind":"pith_short_8","alias_value":"OHUJH2YV","created_at":"2026-07-05T04:40:03.388313+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.02112","citing_title":"An adaptive variance estimator for relative sparsity","ref_index":67,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OHUJH2YVDRMMNTHZGNQ4K62FOE","json":"https://pith.science/pith/OHUJH2YVDRMMNTHZGNQ4K62FOE.json","graph_json":"https://pith.science/api/pith-number/OHUJH2YVDRMMNTHZGNQ4K62FOE/graph.json","events_json":"https://pith.science/api/pith-number/OHUJH2YVDRMMNTHZGNQ4K62FOE/events.json","paper":"https://pith.science/paper/OHUJH2YV"},"agent_actions":{"view_html":"https://pith.science/pith/OHUJH2YVDRMMNTHZGNQ4K62FOE","download_json":"https://pith.science/pith/OHUJH2YVDRMMNTHZGNQ4K62FOE.json","view_paper":"https://pith.science/paper/OHUJH2YV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2207.06269&json=true","fetch_graph":"https://pith.science/api/pith-number/OHUJH2YVDRMMNTHZGNQ4K62FOE/graph.json","fetch_events":"https://pith.science/api/pith-number/OHUJH2YVDRMMNTHZGNQ4K62FOE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OHUJH2YVDRMMNTHZGNQ4K62FOE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OHUJH2YVDRMMNTHZGNQ4K62FOE/action/storage_attestation","attest_author":"https://pith.science/pith/OHUJH2YVDRMMNTHZGNQ4K62FOE/action/author_attestation","sign_citation":"https://pith.science/pith/OHUJH2YVDRMMNTHZGNQ4K62FOE/action/citation_signature","submit_replication":"https://pith.science/pith/OHUJH2YVDRMMNTHZGNQ4K62FOE/action/replication_record"}},"created_at":"2026-07-05T04:40:03.388313+00:00","updated_at":"2026-07-05T04:40:03.388313+00:00"}