{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:MGGLWU5UKLRCFV5TEVT3NKOAOF","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"57a86ccb766ce6ae2057783515ca3c22237070ed30ae285af0b87a6c117dee5f","cross_cats_sorted":["cs.AI","stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ME","submitted_at":"2026-04-26T22:38:23Z","title_canon_sha256":"fa94d9b896e0fd379a1e07a33d6a8614f0b383be1c52d47d800c61fce6fe3ffc"},"schema_version":"1.0","source":{"id":"2604.23904","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.23904","created_at":"2026-07-17T00:20:47Z"},{"alias_kind":"arxiv_version","alias_value":"2604.23904v3","created_at":"2026-07-17T00:20:47Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.23904","created_at":"2026-07-17T00:20:47Z"},{"alias_kind":"pith_short_12","alias_value":"MGGLWU5UKLRC","created_at":"2026-07-17T00:20:47Z"},{"alias_kind":"pith_short_16","alias_value":"MGGLWU5UKLRCFV5T","created_at":"2026-07-17T00:20:47Z"},{"alias_kind":"pith_short_8","alias_value":"MGGLWU5U","created_at":"2026-07-17T00:20:47Z"}],"graph_snapshots":[{"event_id":"sha256:9d74287c85f5b8e68ba22ab2d9797fe81e316a032db59c7bf6ef2639cc9f4891","target":"graph","created_at":"2026-07-17T00:20:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"We show that fully generative tabular synthesizers, including GAN- and LLM-based models, can preserve predictive utility while distorting average treatment effect (ATE) estimates. The failure is structural: ATE preservation requires both a realistic covariate law and an accurate treatment-effect contrast, whereas prediction loss penalizes treatment-effect error only through an overlap-weighted term."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The assumption that separating covariate generation from treatment and outcome modeling in the hybrid framework can be done without introducing new biases or distortions while still producing realistic data for causal inference."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Fully generative synthetic data preserves predictive utility but distorts ATE estimates due to a structural mismatch with prediction loss; a hybrid framework separating covariate generation from causal mechanisms improves fidelity."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Fully generative synthetic tabular data often preserves predictive performance while distorting average treatment effect estimates, because prediction loss only weakly constrains the treatment contrast."}],"snapshot_sha256":"a051beb20319260f171cec5ddd77e4993f25eeefa2f5aee0f9910f0ae8b09d7b"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-21T07:42:17.976056Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T22:36:27.490366Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.23904/integrity.json","findings":[],"snapshot_sha256":"6e4a1533ebe6a991fbab442646f9958553b847137d7ebf6e6ac96604a630884d","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Synthetic tabular data are often evaluated by distributional similarity, privacy distance, or train-on-synthetic-test-on-real predictive performance, but these criteria do not ensure validity for causal inference. We show that fully generative tabular synthesizers, including GAN- and LLM-based models, can preserve predictive utility while distorting average treatment effect (ATE) estimates. The failure is structural: ATE preservation requires both a realistic covariate law and an accurate treatment-effect contrast, whereas prediction loss penalizes treatment-effect error only through an overla","authors_text":"Yichen Xu","cross_cats":["cs.AI","stat.ML"],"headline":"Fully generative synthetic tabular data often preserves predictive performance while distorting average treatment effect estimates, because prediction loss only weakly constrains the treatment contrast.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ME","submitted_at":"2026-04-26T22:38:23Z","title":"Generative Synthetic Data for Causal Inference: Pitfalls, Remedies, and Opportunities"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.23904","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-12T02:09:22.968844Z","id":"afb40d62-40c2-49c0-be88-b14a4f66168e","model_set":{"reader":"grok-4.3"},"one_line_summary":"Fully generative synthetic data preserves predictive utility but distorts ATE estimates due to a structural mismatch with prediction loss; a hybrid framework separating covariate generation from causal mechanisms improves fidelity.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Fully generative synthetic tabular data often preserves predictive performance while distorting average treatment effect estimates, because prediction loss only weakly constrains the treatment contrast.","strongest_claim":"We show that fully generative tabular synthesizers, including GAN- and LLM-based models, can preserve predictive utility while distorting average treatment effect (ATE) estimates. The failure is structural: ATE preservation requires both a realistic covariate law and an accurate treatment-effect contrast, whereas prediction loss penalizes treatment-effect error only through an overlap-weighted term.","weakest_assumption":"The assumption that separating covariate generation from treatment and outcome modeling in the hybrid framework can be done without introducing new biases or distortions while still producing realistic data for causal inference."}},"verdict_id":"afb40d62-40c2-49c0-be88-b14a4f66168e"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:80a4ee9614e18552b9927d148cc8f4381e5a325931ef29291496a2ce6261163a","target":"record","created_at":"2026-07-17T00:20:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"57a86ccb766ce6ae2057783515ca3c22237070ed30ae285af0b87a6c117dee5f","cross_cats_sorted":["cs.AI","stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ME","submitted_at":"2026-04-26T22:38:23Z","title_canon_sha256":"fa94d9b896e0fd379a1e07a33d6a8614f0b383be1c52d47d800c61fce6fe3ffc"},"schema_version":"1.0","source":{"id":"2604.23904","kind":"arxiv","version":3}},"canonical_sha256":"618cbb53b452e222d7b32567b6a9c071766b1b1fc1e7d850ead24eefc45e366b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"618cbb53b452e222d7b32567b6a9c071766b1b1fc1e7d850ead24eefc45e366b","first_computed_at":"2026-07-17T00:20:47.328198Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-17T00:20:47.328198Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"+ClbFpBqhv7VTiI3+opx3z/aWZdOUmLBgH9G/nrbKPdJqAfGDs4x516t2ztDyIKs8HL/xmPQ0kpaoyDCr6M5Aw==","signature_status":"signed_v1","signed_at":"2026-07-17T00:20:47.329116Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.23904","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:80a4ee9614e18552b9927d148cc8f4381e5a325931ef29291496a2ce6261163a","sha256:9d74287c85f5b8e68ba22ab2d9797fe81e316a032db59c7bf6ef2639cc9f4891"],"state_sha256":"61be169c9affa09e7e3094841cc5418d88ccc28132ef307eecf5f95be8737404"}