{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:V2TRBCZY4MTUAPZAD54NBKPC5R","short_pith_number":"pith:V2TRBCZY","schema_version":"1.0","canonical_sha256":"aea7108b38e327403f201f78d0a9e2ec57d96f3fd0b5a00c9151879a04bd42ab","source":{"kind":"arxiv","id":"1905.05824","version":3},"attestation_state":"computed","paper":{"title":"Counterfactual Off-Policy Evaluation with Gumbel-Max Structural Causal Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"David Sontag, Michael Oberst","submitted_at":"2019-05-14T20:18:31Z","abstract_excerpt":"We introduce an off-policy evaluation procedure for highlighting episodes where applying a reinforcement learned (RL) policy is likely to have produced a substantially different outcome than the observed policy. In particular, we introduce a class of structural causal models (SCMs) for generating counterfactual trajectories in finite partially observable Markov Decision Processes (POMDPs). We see this as a useful procedure for off-policy \"debugging\" in high-risk settings (e.g., healthcare); by decomposing the expected difference in reward between the RL and observed policy into specific episod"},"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":"1905.05824","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2019-05-14T20:18:31Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"dd5a0821e3adc4c33f39e03e9168a71ba114944fa61829d2259651ec61f3efe8","abstract_canon_sha256":"cfbac66950568c46e4a12425bd3a531d74d613f7c7361e192fa821434b3dd259"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:19:53.001974Z","signature_b64":"fX6GjR5BOTRPQezG2bT2fPkW0dLtf2oUjZFoGtTbI9UPTx5CkqvO6Qxw6AsUsMNX0DgbQBXlJLmeuquIjMlKAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"aea7108b38e327403f201f78d0a9e2ec57d96f3fd0b5a00c9151879a04bd42ab","last_reissued_at":"2026-07-05T02:19:53.001570Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:19:53.001570Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Counterfactual Off-Policy Evaluation with Gumbel-Max Structural Causal Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"David Sontag, Michael Oberst","submitted_at":"2019-05-14T20:18:31Z","abstract_excerpt":"We introduce an off-policy evaluation procedure for highlighting episodes where applying a reinforcement learned (RL) policy is likely to have produced a substantially different outcome than the observed policy. In particular, we introduce a class of structural causal models (SCMs) for generating counterfactual trajectories in finite partially observable Markov Decision Processes (POMDPs). We see this as a useful procedure for off-policy \"debugging\" in high-risk settings (e.g., healthcare); by decomposing the expected difference in reward between the RL and observed policy into specific episod"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1905.05824","kind":"arxiv","version":3},"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/1905.05824/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":"1905.05824","created_at":"2026-07-05T02:19:53.001626+00:00"},{"alias_kind":"arxiv_version","alias_value":"1905.05824v3","created_at":"2026-07-05T02:19:53.001626+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1905.05824","created_at":"2026-07-05T02:19:53.001626+00:00"},{"alias_kind":"pith_short_12","alias_value":"V2TRBCZY4MTU","created_at":"2026-07-05T02:19:53.001626+00:00"},{"alias_kind":"pith_short_16","alias_value":"V2TRBCZY4MTUAPZA","created_at":"2026-07-05T02:19:53.001626+00:00"},{"alias_kind":"pith_short_8","alias_value":"V2TRBCZY","created_at":"2026-07-05T02:19:53.001626+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2501.12823","citing_title":"To Measure or Not: A Cost-Sensitive, Selective Measuring Environment for Agricultural Management Decisions with Reinforcement Learning","ref_index":38,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/V2TRBCZY4MTUAPZAD54NBKPC5R","json":"https://pith.science/pith/V2TRBCZY4MTUAPZAD54NBKPC5R.json","graph_json":"https://pith.science/api/pith-number/V2TRBCZY4MTUAPZAD54NBKPC5R/graph.json","events_json":"https://pith.science/api/pith-number/V2TRBCZY4MTUAPZAD54NBKPC5R/events.json","paper":"https://pith.science/paper/V2TRBCZY"},"agent_actions":{"view_html":"https://pith.science/pith/V2TRBCZY4MTUAPZAD54NBKPC5R","download_json":"https://pith.science/pith/V2TRBCZY4MTUAPZAD54NBKPC5R.json","view_paper":"https://pith.science/paper/V2TRBCZY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1905.05824&json=true","fetch_graph":"https://pith.science/api/pith-number/V2TRBCZY4MTUAPZAD54NBKPC5R/graph.json","fetch_events":"https://pith.science/api/pith-number/V2TRBCZY4MTUAPZAD54NBKPC5R/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V2TRBCZY4MTUAPZAD54NBKPC5R/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V2TRBCZY4MTUAPZAD54NBKPC5R/action/storage_attestation","attest_author":"https://pith.science/pith/V2TRBCZY4MTUAPZAD54NBKPC5R/action/author_attestation","sign_citation":"https://pith.science/pith/V2TRBCZY4MTUAPZAD54NBKPC5R/action/citation_signature","submit_replication":"https://pith.science/pith/V2TRBCZY4MTUAPZAD54NBKPC5R/action/replication_record"}},"created_at":"2026-07-05T02:19:53.001626+00:00","updated_at":"2026-07-05T02:19:53.001626+00:00"}