{"paper":{"title":"Generative Modeling from Black-box Corruptions via Self-Consistent Stochastic Interpolants","license":"http://creativecommons.org/licenses/by/4.0/","headline":"An iterative procedure with stochastic interpolants learns a transport map that inverts black-box corruptions to generate clean data from corrupted observations alone.","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Chirag Modi, Eric Vanden-Eijnden, Jiequn Han, Joan Bruna","submitted_at":"2025-12-11T17:53:38Z","abstract_excerpt":"Transport-based methods have emerged as a leading paradigm for building generative models from large, clean datasets. However, in many scientific and engineering domains, clean data are often unavailable: instead, we only observe measurements corrupted through a noisy, ill-conditioned channel. A generative model for the original data thus requires solving an inverse problem at the level of distributions. In this work, we introduce a novel approach to this task based on Stochastic Interpolants: we iteratively update a transport map between corrupted and clean data samples using only access to t"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Under appropriate conditions, this iterative procedure converges towards a self-consistent transport map that effectively inverts the corruption channel, thus enabling a generative model for the clean data.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The iterative procedure converges to the desired self-consistent transport map under appropriate (unspecified in abstract) conditions on the corruption channel and data distributions.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SCSI iteratively refines a self-consistent transport map to invert black-box corruptions and enable generative modeling of clean data.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"An iterative procedure with stochastic interpolants learns a transport map that inverts black-box corruptions to generate clean data from corrupted observations alone.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"274414270b7fcb0cb953207e90a878712024d0a68875a54d0b076c2cb2257fd7"},"source":{"id":"2512.10857","kind":"arxiv","version":2},"verdict":{"id":"e2f790ff-9258-4801-b297-32b50b68459d","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T22:59:02.873512Z","strongest_claim":"Under appropriate conditions, this iterative procedure converges towards a self-consistent transport map that effectively inverts the corruption channel, thus enabling a generative model for the clean data.","one_line_summary":"SCSI iteratively refines a self-consistent transport map to invert black-box corruptions and enable generative modeling of clean data.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The iterative procedure converges to the desired self-consistent transport map under appropriate (unspecified in abstract) conditions on the corruption channel and data distributions.","pith_extraction_headline":"An iterative procedure with stochastic interpolants learns a transport map that inverts black-box corruptions to generate clean data from corrupted observations alone."},"references":{"count":18,"sample":[{"doi":"","year":null,"title":"Stochastic Interpolants: A Unifying Framework for Flows and Diffusions","work_id":"c2c7dd8f-fbfb-4591-89ec-9a3a0e6744bd","ref_index":1,"cited_arxiv_id":"2303.08797","is_internal_anchor":true},{"doi":"","year":null,"title":"Stochastic interpolants with data-dependent couplings.arXiv preprint arXiv:2310.03725,","work_id":"9d29254d-1e1b-42c3-ac06-35e6ef5d5d16","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Building normalizing flows with stochastic interpolants","work_id":"85ffbc5d-05cf-43b9-88b7-94664bc37be4","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Nearlyd-linear convergence bounds for diffusion models via stochastic local- ization.CoRR, abs/2308.03686","work_id":"c042e9f5-74ae-4fea-b435-c2920ccb9e2e","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"arXiv preprint arXiv:2209.11215 , year =","work_id":"311398cb-1dd2-4542-88ad-c173f1545e4a","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":18,"snapshot_sha256":"127863a653e86cbe91bfb68002d676a9c7d9f0c38368d7375417b13848c1ab9b","internal_anchors":2},"formal_canon":{"evidence_count":2,"snapshot_sha256":"b639b8e8390900e35dcda75dd5ccfced9ebc04367101d45e95e4f642d8b5c145"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}