{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:FO74R6OUPBUF3QETFNRK354RXO","short_pith_number":"pith:FO74R6OU","schema_version":"1.0","canonical_sha256":"2bbfc8f9d478685dc0932b62adf791bbb42a7426f2b279a5ee0f34ae8ab03e21","source":{"kind":"arxiv","id":"2105.02375","version":1},"attestation_state":"computed","paper":{"title":"A Geometric Analysis of Neural Collapse with Unconstrained Features","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.IT","math.IT","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Chong You, Jeremias Sulam, Jinxin Zhou, Qing Qu, Tianyu Ding, Xiao Li, Zhihui Zhu","submitted_at":"2021-05-06T00:00:50Z","abstract_excerpt":"We provide the first global optimization landscape analysis of $Neural\\;Collapse$ -- an intriguing empirical phenomenon that arises in the last-layer classifiers and features of neural networks during the terminal phase of training. As recently reported by Papyan et al., this phenomenon implies that ($i$) the class means and the last-layer classifiers all collapse to the vertices of a Simplex Equiangular Tight Frame (ETF) up to scaling, and ($ii$) cross-example within-class variability of last-layer activations collapses to zero. We study the problem based on a simplified $unconstrained\\;featu"},"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":"2105.02375","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-05-06T00:00:50Z","cross_cats_sorted":["cs.AI","cs.IT","math.IT","math.OC","stat.ML"],"title_canon_sha256":"1ae398a4d161ac0968a8739e32cf09e1e9230e1d775d518e2bd9c9cbf8c83f7b","abstract_canon_sha256":"ab566cf558e3bac7dced988dcc372679c2896efeadf207ed6f8e4e5995c9bf3c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:38:00.687879Z","signature_b64":"hUKE99FxDV0tfKN5LdzqjcmUW5zPaXrgRm6MESpFIzi1jyYcJpr6DfUrH8kY2xX5qJm33/MBDeNSTJb1VjlTDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2bbfc8f9d478685dc0932b62adf791bbb42a7426f2b279a5ee0f34ae8ab03e21","last_reissued_at":"2026-07-05T02:38:00.687409Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:38:00.687409Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Geometric Analysis of Neural Collapse with Unconstrained Features","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.IT","math.IT","math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Chong You, Jeremias Sulam, Jinxin Zhou, Qing Qu, Tianyu Ding, Xiao Li, Zhihui Zhu","submitted_at":"2021-05-06T00:00:50Z","abstract_excerpt":"We provide the first global optimization landscape analysis of $Neural\\;Collapse$ -- an intriguing empirical phenomenon that arises in the last-layer classifiers and features of neural networks during the terminal phase of training. As recently reported by Papyan et al., this phenomenon implies that ($i$) the class means and the last-layer classifiers all collapse to the vertices of a Simplex Equiangular Tight Frame (ETF) up to scaling, and ($ii$) cross-example within-class variability of last-layer activations collapses to zero. We study the problem based on a simplified $unconstrained\\;featu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2105.02375","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/2105.02375/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":"2105.02375","created_at":"2026-07-05T02:38:00.687467+00:00"},{"alias_kind":"arxiv_version","alias_value":"2105.02375v1","created_at":"2026-07-05T02:38:00.687467+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2105.02375","created_at":"2026-07-05T02:38:00.687467+00:00"},{"alias_kind":"pith_short_12","alias_value":"FO74R6OUPBUF","created_at":"2026-07-05T02:38:00.687467+00:00"},{"alias_kind":"pith_short_16","alias_value":"FO74R6OUPBUF3QET","created_at":"2026-07-05T02:38:00.687467+00:00"},{"alias_kind":"pith_short_8","alias_value":"FO74R6OU","created_at":"2026-07-05T02:38:00.687467+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.09487","citing_title":"Neural Collapse Is Forbidden: Information Floors in Language Models","ref_index":20,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FO74R6OUPBUF3QETFNRK354RXO","json":"https://pith.science/pith/FO74R6OUPBUF3QETFNRK354RXO.json","graph_json":"https://pith.science/api/pith-number/FO74R6OUPBUF3QETFNRK354RXO/graph.json","events_json":"https://pith.science/api/pith-number/FO74R6OUPBUF3QETFNRK354RXO/events.json","paper":"https://pith.science/paper/FO74R6OU"},"agent_actions":{"view_html":"https://pith.science/pith/FO74R6OUPBUF3QETFNRK354RXO","download_json":"https://pith.science/pith/FO74R6OUPBUF3QETFNRK354RXO.json","view_paper":"https://pith.science/paper/FO74R6OU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2105.02375&json=true","fetch_graph":"https://pith.science/api/pith-number/FO74R6OUPBUF3QETFNRK354RXO/graph.json","fetch_events":"https://pith.science/api/pith-number/FO74R6OUPBUF3QETFNRK354RXO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FO74R6OUPBUF3QETFNRK354RXO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FO74R6OUPBUF3QETFNRK354RXO/action/storage_attestation","attest_author":"https://pith.science/pith/FO74R6OUPBUF3QETFNRK354RXO/action/author_attestation","sign_citation":"https://pith.science/pith/FO74R6OUPBUF3QETFNRK354RXO/action/citation_signature","submit_replication":"https://pith.science/pith/FO74R6OUPBUF3QETFNRK354RXO/action/replication_record"}},"created_at":"2026-07-05T02:38:00.687467+00:00","updated_at":"2026-07-05T02:38:00.687467+00:00"}