{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:HFRU4635BD67XKHYVJBAHB5BQK","short_pith_number":"pith:HFRU4635","schema_version":"1.0","canonical_sha256":"39634e7b7d08fdfba8f8aa420387a18298ef71e21dcc8d38864c412a0de3fed2","source":{"kind":"arxiv","id":"2602.10680","version":1},"attestation_state":"computed","paper":{"title":"A solvable high-dimensional model where nonlinear autoencoders learn structure invisible to PCA while test loss misaligns with generalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.LG"],"primary_cat":"stat.ML","authors_text":"C\\'edric Koller, Emanuele Troiani, Jorge Medina Moreira, Lenka Zdeborov\\'a, Lorenzo Bardone, Vicente Conde Mendes, Vittorio Erba","submitted_at":"2026-02-11T09:31:29Z","abstract_excerpt":"Many real-world datasets contain hidden structure that cannot be detected by simple linear correlations between input features. For example, latent factors may influence the data in a coordinated way, even though their effect is invisible to covariance-based methods such as PCA. In practice, nonlinear neural networks often succeed in extracting such hidden structure in unsupervised and self-supervised learning. However, constructing a minimal high-dimensional model where this advantage can be rigorously analyzed has remained an open theoretical challenge. We introduce a tractable high-dimensio"},"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":"2602.10680","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-02-11T09:31:29Z","cross_cats_sorted":["cond-mat.dis-nn","cs.LG"],"title_canon_sha256":"d4d1a3b06f9c92f07e00ecca90279bd7113757b55c7bf3608e59ac760c69f4be","abstract_canon_sha256":"0d3abf5023975c1699439316251c72d516f04d7920ea46d550338127bb37994b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-08T01:03:58.506565Z","signature_b64":"MEk5ZlPEZ7IuFAPk8h9BWWsP7Q4TiXCmXzoHQGl9n9Ks3lsYam/oyp6J4d8LmrK62LT20gzdV15mXfg//6EQAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"39634e7b7d08fdfba8f8aa420387a18298ef71e21dcc8d38864c412a0de3fed2","last_reissued_at":"2026-06-08T01:03:58.505639Z","signature_status":"signed_v1","first_computed_at":"2026-06-08T01:03:58.505639Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A solvable high-dimensional model where nonlinear autoencoders learn structure invisible to PCA while test loss misaligns with generalization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.LG"],"primary_cat":"stat.ML","authors_text":"C\\'edric Koller, Emanuele Troiani, Jorge Medina Moreira, Lenka Zdeborov\\'a, Lorenzo Bardone, Vicente Conde Mendes, Vittorio Erba","submitted_at":"2026-02-11T09:31:29Z","abstract_excerpt":"Many real-world datasets contain hidden structure that cannot be detected by simple linear correlations between input features. For example, latent factors may influence the data in a coordinated way, even though their effect is invisible to covariance-based methods such as PCA. In practice, nonlinear neural networks often succeed in extracting such hidden structure in unsupervised and self-supervised learning. However, constructing a minimal high-dimensional model where this advantage can be rigorously analyzed has remained an open theoretical challenge. We introduce a tractable high-dimensio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.10680","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/2602.10680/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":"2602.10680","created_at":"2026-06-08T01:03:58.505767+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.10680v1","created_at":"2026-06-08T01:03:58.505767+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.10680","created_at":"2026-06-08T01:03:58.505767+00:00"},{"alias_kind":"pith_short_12","alias_value":"HFRU4635BD67","created_at":"2026-06-08T01:03:58.505767+00:00"},{"alias_kind":"pith_short_16","alias_value":"HFRU4635BD67XKHY","created_at":"2026-06-08T01:03:58.505767+00:00"},{"alias_kind":"pith_short_8","alias_value":"HFRU4635","created_at":"2026-06-08T01:03:58.505767+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"2603.12901","citing_title":"A theory of learning data statistics in diffusion models, from easy to hard","ref_index":37,"is_internal_anchor":true},{"citing_arxiv_id":"2605.21402","citing_title":"Memorisation, convergence and generalisation in generative models","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2605.16913","citing_title":"A Fourier perspective on the learning dynamics of neural networks: from sample complexities to mechanistic insights","ref_index":54,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06367","citing_title":"The Interplay of Data Structure and Imbalance in the Learning Dynamics of Diffusion Models","ref_index":35,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/HFRU4635BD67XKHYVJBAHB5BQK","json":"https://pith.science/pith/HFRU4635BD67XKHYVJBAHB5BQK.json","graph_json":"https://pith.science/api/pith-number/HFRU4635BD67XKHYVJBAHB5BQK/graph.json","events_json":"https://pith.science/api/pith-number/HFRU4635BD67XKHYVJBAHB5BQK/events.json","paper":"https://pith.science/paper/HFRU4635"},"agent_actions":{"view_html":"https://pith.science/pith/HFRU4635BD67XKHYVJBAHB5BQK","download_json":"https://pith.science/pith/HFRU4635BD67XKHYVJBAHB5BQK.json","view_paper":"https://pith.science/paper/HFRU4635","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.10680&json=true","fetch_graph":"https://pith.science/api/pith-number/HFRU4635BD67XKHYVJBAHB5BQK/graph.json","fetch_events":"https://pith.science/api/pith-number/HFRU4635BD67XKHYVJBAHB5BQK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HFRU4635BD67XKHYVJBAHB5BQK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HFRU4635BD67XKHYVJBAHB5BQK/action/storage_attestation","attest_author":"https://pith.science/pith/HFRU4635BD67XKHYVJBAHB5BQK/action/author_attestation","sign_citation":"https://pith.science/pith/HFRU4635BD67XKHYVJBAHB5BQK/action/citation_signature","submit_replication":"https://pith.science/pith/HFRU4635BD67XKHYVJBAHB5BQK/action/replication_record"}},"created_at":"2026-06-08T01:03:58.505767+00:00","updated_at":"2026-06-08T01:03:58.505767+00:00"}