{"paper":{"title":"CausalReasoningBenchmark: A Real-World Benchmark for Disentangled Evaluation of Causal Identification and Estimation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"A benchmark of 173 real-world queries scores causal identification and numerical estimation separately to diagnose AI failures in causal analysis.","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Ayush Sawarni, Jiyuan Tan, Vasilis Syrgkanis","submitted_at":"2026-02-24T05:44:25Z","abstract_excerpt":"Many benchmarks for automated causal inference evaluate a system's performance based on a single numerical output, such as an Average Treatment Effect (ATE). This approach conflates two distinct steps in causal analysis: identification - formulating a valid research design under stated assumptions - and estimation - implementing that design numerically on finite data. We introduce CausalReasoningBenchmark, a benchmark of 173 queries across 132 real-world datasets, curated from 79 peer-reviewed research papers and three widely-used causal-inference textbooks. For each query a system must produc"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"By scoring these two components separately, our benchmark enables granular diagnosis: it distinguishes failures in causal reasoning from errors in numerical execution.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The ground-truth identification specifications and estimates extracted from the 79 source papers and three textbooks are accurate and complete enough to serve as reliable labels for the 173 queries.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CausalReasoningBenchmark supplies 173 real-world queries that separately grade causal identification specifications and point estimates to expose distinct failure modes in automated causal systems.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"A benchmark of 173 real-world queries scores causal identification and numerical estimation separately to diagnose AI failures in causal analysis.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ece80de0c697663bcf126a20ea98117408a338d17cb1234e681fdd0f49f8ab00"},"source":{"id":"2602.20571","kind":"arxiv","version":2},"verdict":{"id":"209e0355-0021-474d-a91c-02e3945eea3a","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T20:32:29.922655Z","strongest_claim":"By scoring these two components separately, our benchmark enables granular diagnosis: it distinguishes failures in causal reasoning from errors in numerical execution.","one_line_summary":"CausalReasoningBenchmark supplies 173 real-world queries that separately grade causal identification specifications and point estimates to expose distinct failure modes in automated causal systems.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The ground-truth identification specifications and estimates extracted from the 79 source papers and three textbooks are accurate and complete enough to serve as reliable labels for the 173 queries.","pith_extraction_headline":"A benchmark of 173 real-world queries scores causal identification and numerical estimation separately to diagnose AI failures in causal analysis."},"references":{"count":95,"sample":[{"doi":"10.1086/681718","year":2015,"title":"Incumbency disadvantage under electoral rules with intraparty competition: Evidence from japan.The Journal of Politics, 2015","work_id":"da92bb41-e05e-4869-95dc-42aace3bbc2b","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.48550/arxiv.2508.10581","year":2025,"title":"Technical report: Facilitating the adoption of causal inference methods through LLM-empowered co-pilot.arXiv preprint arXiv:2508.10581, 2025","work_id":"840b547f-b136-40f7-808d-1bd87148568e","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1086/715255","year":2022,"title":"How does armed conflict shape investment? evidence from the mining sector.The Journal of Politics, 2022","work_id":"56263026-baa4-418c-b189-ce45abc9dd1b","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1017/s002238161300145x","year":2014,"title":"Taylor C. Boas, F. Daniel Hidalgo, and Neal P. Richardson. The spoils of victory: Campaign donations and government contracts in brazil.The Journal of Politics, 2014. doi: 10.1017/s002238161300145x. U","work_id":"15a5646d-4c43-4aa0-a6b7-8ba20320ff1a","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1111/ajps.12228","year":2015,"title":"Broockman and Timothy J","work_id":"db0bdcd2-b7f0-4fcc-8bac-a557b16c8582","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":95,"snapshot_sha256":"ec469b7b7ead7457fe7223ef4c02c0036e43b33e8846b55930d607b527118115","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"}