{"paper":{"title":"Information-Preserving Domain Transfer with Unlabeled Data in Misspecified Simulation-Based Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"SPIN improves posterior inference in misspecified simulation-based inference by using information-preserving domain transfer with unlabeled real-world data.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Eunho Jeong, Hyeonjin Kim, Joon Jang, Kyu Sung Choi","submitted_at":"2026-05-07T04:06:53Z","abstract_excerpt":"Simulation-based inference (SBI) provides amortized Bayesian parameter inference from simulator-generated data without requiring explicit likelihood evaluation. Its reliability can degrade under model misspecification, where real-world observations are not well represented by the simulator used for training. Existing methods using unlabeled real-world data often align simulated and real-world data distributions, but marginal alignment alone does not directly preserve parameter-relevant information needed for posterior inference. We propose SPIN, an SBI framework with parameter-relevant informa"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"SPIN improves real-world posterior inference in misspecified SBI by translating labeled simulator observations toward the real-world domain and back while using original labels to preserve parameter-relevant mutual information, with gains becoming clearer as misspecification increases.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that the learned real-to-simulator transport map preserves the mutual information between observations and parameters sufficiently well that the downstream SBI posterior remains accurate, even though no real-world parameter labels are available during training or testing.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SPIN performs bidirectional domain transfer in SBI to retain parameter mutual information from unlabeled real observations, improving real-world posterior inference under increasing misspecification.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"SPIN improves posterior inference in misspecified simulation-based inference by using information-preserving domain transfer with unlabeled real-world data.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"874c29b0f1c3d84ea8555b54d8c19ae145363a349e498757174b3c4cd305cfdc"},"source":{"id":"2605.05652","kind":"arxiv","version":2},"verdict":{"id":"92768577-fc78-4c70-97aa-b67e36e35b23","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T17:12:14.128838Z","strongest_claim":"SPIN improves real-world posterior inference in misspecified SBI by translating labeled simulator observations toward the real-world domain and back while using original labels to preserve parameter-relevant mutual information, with gains becoming clearer as misspecification increases.","one_line_summary":"SPIN performs bidirectional domain transfer in SBI to retain parameter mutual information from unlabeled real observations, improving real-world posterior inference under increasing misspecification.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that the learned real-to-simulator transport map preserves the mutual information between observations and parameters sufficiently well that the downstream SBI posterior remains accurate, even though no real-world parameter labels are available during training or testing.","pith_extraction_headline":"SPIN improves posterior inference in misspecified simulation-based inference by using information-preserving domain transfer with unlabeled real-world data."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.05652/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T20:01:19.826470Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T13:23:23.669292Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"9a8a1c6568d632d32109578b3f211ba9b31c105d4e51990fe3f3a89d8c4fea0b"},"references":{"count":56,"sample":[{"doi":"10.1073/pnas.1912789117","year":2020,"title":"The frontier of simulation-based inference","work_id":"b1065441-311d-4f7f-b3a4-83827c5a3904","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Simulation-based inference: A practical guide.arXiv preprint arXiv:2508.12939,","work_id":"8c93082f-4e79-4893-9100-4b3691c28db4","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2016,"title":"Fastϵ-free inference of simulation models with Bayesian conditional density estimation","work_id":"c2afcfa1-cdd2-484f-b22c-99090d08a1cb","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Greenberg, Marcel Nonnenmacher, and Jakob H","work_id":"f083a9ca-94c1-4060-b545-c70e2cb8f1bb","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"URLhttps://proceedings.mlr.press/v97/greenberg19a.html","work_id":"4c92d9b3-a58d-4320-b847-5873e339d2a4","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":56,"snapshot_sha256":"8f610ef5cabc4e3db8d2041886ac05131e4a6155049e89dfbe7133b31edbf8bb","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"f08d65832ead454235ed0aae61b602d3ac9f6d7287ec3576d65480d3285004f8"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}