{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:AK6DJQZIBEOOA4WRB3PALKEXNP","short_pith_number":"pith:AK6DJQZI","schema_version":"1.0","canonical_sha256":"02bc34c328091ce072d10ede05a8976bc01b58b6b99f73165656b4465a14e042","source":{"kind":"arxiv","id":"2605.15353","version":1},"attestation_state":"computed","paper":{"title":"PACER: Acyclic Causal Discovery from Large-Scale Interventional Data","license":"http://creativecommons.org/licenses/by/4.0/","headline":"PACER guarantees acyclicity by jointly modeling variable permutations and edge probabilities for direct optimization over valid causal structures.","cross_cats":["cs.AI","q-bio.MN","q-bio.QM"],"primary_cat":"cs.LG","authors_text":"Artyom Gadetsky, Ivo Alexander Ban, Maria Brbi\\'c, Nikita Doikov, Ramon Vi\\~nas Torn\\'e, S\\'ilvia F\\`abregas Salazar, Soyon Park","submitted_at":"2026-05-14T19:26:53Z","abstract_excerpt":"Inferring the structure of directed acyclic graphs (DAGs) from data is a central challenge in causal discovery, particularly in modern high-dimensional settings where large-scale interventional data are increasingly available. While interventional data can improve identifiability, existing methods remain limited by soft acyclicity constraints, leading to optimization over invalid cyclic graphs, numerical instability, and reduced scalability. We introduce PACER (Perturbation-driven Acyclic Causal Edge Recovery), a scalable framework for causal discovery that guarantees acyclicity by constructio"},"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":true,"formal_links_present":true},"canonical_record":{"source":{"id":"2605.15353","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T19:26:53Z","cross_cats_sorted":["cs.AI","q-bio.MN","q-bio.QM"],"title_canon_sha256":"b16b3bf2bded30a7eeb991f3455e34f3400093606f5999cd77d0baf548d2ae54","abstract_canon_sha256":"8674d5e0bf2d1cb9aa8b510fab3c169377fed5d8f11bd3bd6df2d2f3066ce6a9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:00:54.022929Z","signature_b64":"RIwmu9g8WOEEuRLbgTH6wMLItHLljwG+NLDjqvRav3vEyDRNC6/RM2a5VSgg1kJGl0/p+u+Xa9jmVPKWulY4CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"02bc34c328091ce072d10ede05a8976bc01b58b6b99f73165656b4465a14e042","last_reissued_at":"2026-05-20T00:00:54.022076Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:00:54.022076Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PACER: Acyclic Causal Discovery from Large-Scale Interventional Data","license":"http://creativecommons.org/licenses/by/4.0/","headline":"PACER guarantees acyclicity by jointly modeling variable permutations and edge probabilities for direct optimization over valid causal structures.","cross_cats":["cs.AI","q-bio.MN","q-bio.QM"],"primary_cat":"cs.LG","authors_text":"Artyom Gadetsky, Ivo Alexander Ban, Maria Brbi\\'c, Nikita Doikov, Ramon Vi\\~nas Torn\\'e, S\\'ilvia F\\`abregas Salazar, Soyon Park","submitted_at":"2026-05-14T19:26:53Z","abstract_excerpt":"Inferring the structure of directed acyclic graphs (DAGs) from data is a central challenge in causal discovery, particularly in modern high-dimensional settings where large-scale interventional data are increasingly available. While interventional data can improve identifiability, existing methods remain limited by soft acyclicity constraints, leading to optimization over invalid cyclic graphs, numerical instability, and reduced scalability. We introduce PACER (Perturbation-driven Acyclic Causal Edge Recovery), a scalable framework for causal discovery that guarantees acyclicity by constructio"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"PACER guarantees acyclicity by construction through a joint model of variable permutations and edge probabilities, enabling direct optimization over valid causal structures and yielding up to two orders of magnitude speedups over penalty-based approaches while matching or exceeding state-of-the-art on protein signaling and large-scale genetic perturbation benchmarks.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The parameterization over permutations and edge probabilities is assumed to efficiently cover the space of all DAGs without introducing bias or missing high-probability structures, particularly when incorporating structural prior knowledge or flexible conditional density models; this enters in the description of the joint model and the optimization procedure.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"PACER guarantees acyclicity by construction for causal discovery by jointly modeling permutations and edge probabilities, supporting unified treatment of observational and interventional data with closed-form expressions for linear-Gaussian cases.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"PACER guarantees acyclicity by jointly modeling variable permutations and edge probabilities for direct optimization over valid causal structures.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"9e1805093284410b4df9860b49601108b512d13479c22b37847272187096ca33"},"source":{"id":"2605.15353","kind":"arxiv","version":1},"verdict":{"id":"8cd45bc5-80a2-4f21-a466-fa31416a5b4e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T15:39:11.264185Z","strongest_claim":"PACER guarantees acyclicity by construction through a joint model of variable permutations and edge probabilities, enabling direct optimization over valid causal structures and yielding up to two orders of magnitude speedups over penalty-based approaches while matching or exceeding state-of-the-art on protein signaling and large-scale genetic perturbation benchmarks.","one_line_summary":"PACER guarantees acyclicity by construction for causal discovery by jointly modeling permutations and edge probabilities, supporting unified treatment of observational and interventional data with closed-form expressions for linear-Gaussian cases.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The parameterization over permutations and edge probabilities is assumed to efficiently cover the space of all DAGs without introducing bias or missing high-probability structures, particularly when incorporating structural prior knowledge or flexible conditional density models; this enters in the description of the joint model and the optimization procedure.","pith_extraction_headline":"PACER guarantees acyclicity by jointly modeling variable permutations and edge probabilities for direct optimization over valid causal structures."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.15353/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T16:01:18.094432Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T15:53:42.814272Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T14:21:54.200921Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T13:33:22.747800Z","status":"skipped","version":"1.0.0","findings_count":0}],"snapshot_sha256":"640c47e4605a2fa775cee355839655dbdbb1c05271f0c559a29a96cc91c30c32"},"references":{"count":12,"sample":[{"doi":"","year":2024,"title":"Badia-i Mompel, P., Casals-Franch, R., Wessels, L., M¨uller- Dott, S., Trimbour, R., Yang, Y ., Ramirez Flores, R. O., and Saez-Rodriguez, J. Comparison and evaluation of methods to infer gene regulat","work_id":"6bdc5e09-1c71-474b-9901-5b02ba30ed41","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2022,"title":"orders of magnitude larger and grows strongly with N (≈90 for N=500 vs","work_id":"10fb4537-959f-4b2a-96c1-8a50c9730c92","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2003,"title":"To complete the proof, we use the formulas,Var(aij) =E[a ij](1−E[aij]) and Cov(aij, akj) =E[a ij]E[akj] eθj eθi+eθj +eθk , that were established in Theorem C.4","work_id":"0dfe8171-cadf-4230-b2fb-fc07566b1fbf","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2018,"title":"NOTEARS (Zheng et al., 2018), NOTEARS-LR (Fang et al., 2023), DCDI (Brouillard et al.,","work_id":"81886afb-3b2f-44d2-96ab-fc0010ec4ae5","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Baselines implementations.GS, GES, GIES, IAMB, MMPC, GRaSP, BOSS, and LiNGAM are benchmarked using the implementation of the Causal Discovery Toolbox (Kalainathan et al., 2020)","work_id":"968ec0a1-7b10-46f5-8b3c-231ed13fc186","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":12,"snapshot_sha256":"716f2110c5014a2c94aebc2b2e2926c13e6fad2e31cba63594b67974e5fedbb9","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"a7a2169142545f9382e7703e8b2f6880de0842b0aac6c72df1774776e827494f"},"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":"2605.15353","created_at":"2026-05-20T00:00:54.022223+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.15353v1","created_at":"2026-05-20T00:00:54.022223+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.15353","created_at":"2026-05-20T00:00:54.022223+00:00"},{"alias_kind":"pith_short_12","alias_value":"AK6DJQZIBEOO","created_at":"2026-05-20T00:00:54.022223+00:00"},{"alias_kind":"pith_short_16","alias_value":"AK6DJQZIBEOOA4WR","created_at":"2026-05-20T00:00:54.022223+00:00"},{"alias_kind":"pith_short_8","alias_value":"AK6DJQZI","created_at":"2026-05-20T00:00:54.022223+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":2,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AK6DJQZIBEOOA4WRB3PALKEXNP","json":"https://pith.science/pith/AK6DJQZIBEOOA4WRB3PALKEXNP.json","graph_json":"https://pith.science/api/pith-number/AK6DJQZIBEOOA4WRB3PALKEXNP/graph.json","events_json":"https://pith.science/api/pith-number/AK6DJQZIBEOOA4WRB3PALKEXNP/events.json","paper":"https://pith.science/paper/AK6DJQZI"},"agent_actions":{"view_html":"https://pith.science/pith/AK6DJQZIBEOOA4WRB3PALKEXNP","download_json":"https://pith.science/pith/AK6DJQZIBEOOA4WRB3PALKEXNP.json","view_paper":"https://pith.science/paper/AK6DJQZI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.15353&json=true","fetch_graph":"https://pith.science/api/pith-number/AK6DJQZIBEOOA4WRB3PALKEXNP/graph.json","fetch_events":"https://pith.science/api/pith-number/AK6DJQZIBEOOA4WRB3PALKEXNP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AK6DJQZIBEOOA4WRB3PALKEXNP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AK6DJQZIBEOOA4WRB3PALKEXNP/action/storage_attestation","attest_author":"https://pith.science/pith/AK6DJQZIBEOOA4WRB3PALKEXNP/action/author_attestation","sign_citation":"https://pith.science/pith/AK6DJQZIBEOOA4WRB3PALKEXNP/action/citation_signature","submit_replication":"https://pith.science/pith/AK6DJQZIBEOOA4WRB3PALKEXNP/action/replication_record"}},"created_at":"2026-05-20T00:00:54.022223+00:00","updated_at":"2026-05-20T00:00:54.022223+00:00"}