{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:RBJVYHNPP4KT73OIEDDWDRQYHH","short_pith_number":"pith:RBJVYHNP","schema_version":"1.0","canonical_sha256":"88535c1daf7f153fedc820c761c61839e22bda00564452b8ac50df5f340b10bb","source":{"kind":"arxiv","id":"2508.07581","version":1},"attestation_state":"computed","paper":{"title":"When and how can inexact generative models still sample from the data manifold?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.DS","math.PR"],"primary_cat":"cs.LG","authors_text":"Adriaan de Clercq, Nisha Chandramoorthy","submitted_at":"2025-08-11T03:24:34Z","abstract_excerpt":"A curious phenomenon observed in some dynamical generative models is the following: despite learning errors in the score function or the drift vector field, the generated samples appear to shift \\emph{along} the support of the data distribution but not \\emph{away} from it. In this work, we investigate this phenomenon of \\emph{robustness of the support} by taking a dynamical systems approach on the generating stochastic/deterministic process. Our perturbation analysis of the probability flow reveals that infinitesimal learning errors cause the predicted density to be different from the target d"},"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":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2508.07581","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-08-11T03:24:34Z","cross_cats_sorted":["math.DS","math.PR"],"title_canon_sha256":"931556cdef8fe448e15413a6e36a4075891b285af63347a72e8e66788537b0a5","abstract_canon_sha256":"080ebfd3e43799b3ec625f640447d8e81c1badb5c9220d7bdf7754c426ad8098"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:51:48.680532Z","signature_b64":"083akIJiVqts+Rk038wCsD39hQ9F2BdpGkeHQfSWUubgYGYWVKJlXU/WdJXZXUF1Vxg20fYfCjPwZcBKghabCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"88535c1daf7f153fedc820c761c61839e22bda00564452b8ac50df5f340b10bb","last_reissued_at":"2026-07-05T11:51:48.679790Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:51:48.679790Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When and how can inexact generative models still sample from the data manifold?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.DS","math.PR"],"primary_cat":"cs.LG","authors_text":"Adriaan de Clercq, Nisha Chandramoorthy","submitted_at":"2025-08-11T03:24:34Z","abstract_excerpt":"A curious phenomenon observed in some dynamical generative models is the following: despite learning errors in the score function or the drift vector field, the generated samples appear to shift \\emph{along} the support of the data distribution but not \\emph{away} from it. In this work, we investigate this phenomenon of \\emph{robustness of the support} by taking a dynamical systems approach on the generating stochastic/deterministic process. Our perturbation analysis of the probability flow reveals that infinitesimal learning errors cause the predicted density to be different from the target d"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.07581","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2508.07581/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2508.07581","created_at":"2026-07-05T11:51:48.679881+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.07581v1","created_at":"2026-07-05T11:51:48.679881+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.07581","created_at":"2026-07-05T11:51:48.679881+00:00"},{"alias_kind":"pith_short_12","alias_value":"RBJVYHNPP4KT","created_at":"2026-07-05T11:51:48.679881+00:00"},{"alias_kind":"pith_short_16","alias_value":"RBJVYHNPP4KT73OI","created_at":"2026-07-05T11:51:48.679881+00:00"},{"alias_kind":"pith_short_8","alias_value":"RBJVYHNP","created_at":"2026-07-05T11:51:48.679881+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.20235","citing_title":"Provably Learning Diffusion Models under the Manifold Hypothesis: Collapse and Refine","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RBJVYHNPP4KT73OIEDDWDRQYHH","json":"https://pith.science/pith/RBJVYHNPP4KT73OIEDDWDRQYHH.json","graph_json":"https://pith.science/api/pith-number/RBJVYHNPP4KT73OIEDDWDRQYHH/graph.json","events_json":"https://pith.science/api/pith-number/RBJVYHNPP4KT73OIEDDWDRQYHH/events.json","paper":"https://pith.science/paper/RBJVYHNP"},"agent_actions":{"view_html":"https://pith.science/pith/RBJVYHNPP4KT73OIEDDWDRQYHH","download_json":"https://pith.science/pith/RBJVYHNPP4KT73OIEDDWDRQYHH.json","view_paper":"https://pith.science/paper/RBJVYHNP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.07581&json=true","fetch_graph":"https://pith.science/api/pith-number/RBJVYHNPP4KT73OIEDDWDRQYHH/graph.json","fetch_events":"https://pith.science/api/pith-number/RBJVYHNPP4KT73OIEDDWDRQYHH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RBJVYHNPP4KT73OIEDDWDRQYHH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RBJVYHNPP4KT73OIEDDWDRQYHH/action/storage_attestation","attest_author":"https://pith.science/pith/RBJVYHNPP4KT73OIEDDWDRQYHH/action/author_attestation","sign_citation":"https://pith.science/pith/RBJVYHNPP4KT73OIEDDWDRQYHH/action/citation_signature","submit_replication":"https://pith.science/pith/RBJVYHNPP4KT73OIEDDWDRQYHH/action/replication_record"}},"created_at":"2026-07-05T11:51:48.679881+00:00","updated_at":"2026-07-05T11:51:48.679881+00:00"}