{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:XRLATOKUTCRVZG5OP32RBC7QZT","short_pith_number":"pith:XRLATOKU","schema_version":"1.0","canonical_sha256":"bc5609b95498a35c9bae7ef5108bf0cce9be137761d892fb31273dc55dbc8200","source":{"kind":"arxiv","id":"2106.02073","version":4},"attestation_state":"computed","paper":{"title":"Neural Collapse Under MSE Loss: Proximity to and Dynamics on the Central Path","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","math.DG","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"David L. Donoho, Vardan Papyan, X.Y. Han","submitted_at":"2021-06-03T18:31:41Z","abstract_excerpt":"The recently discovered Neural Collapse (NC) phenomenon occurs pervasively in today's deep net training paradigm of driving cross-entropy (CE) loss towards zero. During NC, last-layer features collapse to their class-means, both classifiers and class-means collapse to the same Simplex Equiangular Tight Frame, and classifier behavior collapses to the nearest-class-mean decision rule. Recent works demonstrated that deep nets trained with mean squared error (MSE) loss perform comparably to those trained with CE. As a preliminary, we empirically establish that NC emerges in such MSE-trained deep n"},"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":"2106.02073","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2021-06-03T18:31:41Z","cross_cats_sorted":["cs.AI","math.DG","math.OC","stat.ML"],"title_canon_sha256":"251c85de37f491a8d7407cf45f320c4f7c201e1d9f9d7f13c2efadf30f4db62a","abstract_canon_sha256":"4c50363b58de9a962f2d7038a276a79aab01ee2f862e9725801a5977ed879f2f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:21:25.985025Z","signature_b64":"5paE2PNzKQFdXtY1xzCLN+KxwxPmAdGTSq/+GaN6+5k2aa4emCgyUtuFWY30CrEwI6/OJzmrUApUw24hU/SqBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bc5609b95498a35c9bae7ef5108bf0cce9be137761d892fb31273dc55dbc8200","last_reissued_at":"2026-07-05T04:21:25.984619Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:21:25.984619Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Neural Collapse Under MSE Loss: Proximity to and Dynamics on the Central Path","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI","math.DG","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"David L. Donoho, Vardan Papyan, X.Y. Han","submitted_at":"2021-06-03T18:31:41Z","abstract_excerpt":"The recently discovered Neural Collapse (NC) phenomenon occurs pervasively in today's deep net training paradigm of driving cross-entropy (CE) loss towards zero. During NC, last-layer features collapse to their class-means, both classifiers and class-means collapse to the same Simplex Equiangular Tight Frame, and classifier behavior collapses to the nearest-class-mean decision rule. Recent works demonstrated that deep nets trained with mean squared error (MSE) loss perform comparably to those trained with CE. As a preliminary, we empirically establish that NC emerges in such MSE-trained deep n"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.02073","kind":"arxiv","version":4},"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/2106.02073/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":"2106.02073","created_at":"2026-07-05T04:21:25.984675+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.02073v4","created_at":"2026-07-05T04:21:25.984675+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.02073","created_at":"2026-07-05T04:21:25.984675+00:00"},{"alias_kind":"pith_short_12","alias_value":"XRLATOKUTCRV","created_at":"2026-07-05T04:21:25.984675+00:00"},{"alias_kind":"pith_short_16","alias_value":"XRLATOKUTCRVZG5O","created_at":"2026-07-05T04:21:25.984675+00:00"},{"alias_kind":"pith_short_8","alias_value":"XRLATOKU","created_at":"2026-07-05T04:21:25.984675+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.30524","citing_title":"Representation Collapse in Sequential Post-Training of Large Language Models","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2505.11771","citing_title":"Residual Feature Integration is Sufficient to Prevent Negative Transfer","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15571","citing_title":"MaxSketch: Robust Distinct Counting in Streams via Random Projections","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2510.01105","citing_title":"Geometric Analysis of Neural Regression Collapse via Intrinsic Dimension","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21691","citing_title":"There Will Be a Scientific Theory of Deep Learning","ref_index":126,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XRLATOKUTCRVZG5OP32RBC7QZT","json":"https://pith.science/pith/XRLATOKUTCRVZG5OP32RBC7QZT.json","graph_json":"https://pith.science/api/pith-number/XRLATOKUTCRVZG5OP32RBC7QZT/graph.json","events_json":"https://pith.science/api/pith-number/XRLATOKUTCRVZG5OP32RBC7QZT/events.json","paper":"https://pith.science/paper/XRLATOKU"},"agent_actions":{"view_html":"https://pith.science/pith/XRLATOKUTCRVZG5OP32RBC7QZT","download_json":"https://pith.science/pith/XRLATOKUTCRVZG5OP32RBC7QZT.json","view_paper":"https://pith.science/paper/XRLATOKU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.02073&json=true","fetch_graph":"https://pith.science/api/pith-number/XRLATOKUTCRVZG5OP32RBC7QZT/graph.json","fetch_events":"https://pith.science/api/pith-number/XRLATOKUTCRVZG5OP32RBC7QZT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XRLATOKUTCRVZG5OP32RBC7QZT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XRLATOKUTCRVZG5OP32RBC7QZT/action/storage_attestation","attest_author":"https://pith.science/pith/XRLATOKUTCRVZG5OP32RBC7QZT/action/author_attestation","sign_citation":"https://pith.science/pith/XRLATOKUTCRVZG5OP32RBC7QZT/action/citation_signature","submit_replication":"https://pith.science/pith/XRLATOKUTCRVZG5OP32RBC7QZT/action/replication_record"}},"created_at":"2026-07-05T04:21:25.984675+00:00","updated_at":"2026-07-05T04:21:25.984675+00:00"}