{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XZDXCK6WABNOITK5ZSVKRRDCN2","short_pith_number":"pith:XZDXCK6W","schema_version":"1.0","canonical_sha256":"be47712bd6005ae44d5dccaaa8c4626ea92d61127284748c250bcec1888e7745","source":{"kind":"arxiv","id":"2509.24914","version":2},"attestation_state":"computed","paper":{"title":"Single-Head Attention in High Dimensions: A Theory of Generalization, Weights Spectra, and Scaling Laws","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.IT","cs.LG","math.IT"],"primary_cat":"stat.ML","authors_text":"Emanuele Troiani, Fabrizio Boncoraglio, Florent Krzakala, Lenka Zdeborov\\'a, Vittorio Erba, Yizhou Xu","submitted_at":"2025-09-29T15:19:31Z","abstract_excerpt":"Trained attention layers exhibit striking and reproducible spectral structure of the weights, including low-rank collapse, bulk deformation, and isolated spectral outliers, yet the origin of these phenomena and their implications for generalization remain poorly understood. We study empirical risk minimization in a single-head tied-attention layer trained on synthetic high-dimensional sequence tasks generated from the attention-indexed model. Using tools from random matrix theory, spin-glass theory, and approximate message passing, we obtain an exact high-dimensional characterization of traini"},"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":"2509.24914","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2025-09-29T15:19:31Z","cross_cats_sorted":["cond-mat.dis-nn","cs.IT","cs.LG","math.IT"],"title_canon_sha256":"efdf7c679a9d430922df9497e75dcc773afd89eb89f58ca955000ffc74975c94","abstract_canon_sha256":"add30c3142fdceddb7468fcad03299387ef34802b5bf49094a1d3532cef79013"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-08T01:03:48.815093Z","signature_b64":"MluI46tWBRI3GTxK2+WoZVneAK1hXRclBzweyaJlrZtO9odCX/KM+LdD8TmRjNE5ZMBBQN4heLYNwDfZB4UZDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"be47712bd6005ae44d5dccaaa8c4626ea92d61127284748c250bcec1888e7745","last_reissued_at":"2026-06-08T01:03:48.814133Z","signature_status":"signed_v1","first_computed_at":"2026-06-08T01:03:48.814133Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Single-Head Attention in High Dimensions: A Theory of Generalization, Weights Spectra, and Scaling Laws","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.IT","cs.LG","math.IT"],"primary_cat":"stat.ML","authors_text":"Emanuele Troiani, Fabrizio Boncoraglio, Florent Krzakala, Lenka Zdeborov\\'a, Vittorio Erba, Yizhou Xu","submitted_at":"2025-09-29T15:19:31Z","abstract_excerpt":"Trained attention layers exhibit striking and reproducible spectral structure of the weights, including low-rank collapse, bulk deformation, and isolated spectral outliers, yet the origin of these phenomena and their implications for generalization remain poorly understood. We study empirical risk minimization in a single-head tied-attention layer trained on synthetic high-dimensional sequence tasks generated from the attention-indexed model. Using tools from random matrix theory, spin-glass theory, and approximate message passing, we obtain an exact high-dimensional characterization of traini"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.24914","kind":"arxiv","version":2},"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/2509.24914/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":"2509.24914","created_at":"2026-06-08T01:03:48.814279+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.24914v2","created_at":"2026-06-08T01:03:48.814279+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.24914","created_at":"2026-06-08T01:03:48.814279+00:00"},{"alias_kind":"pith_short_12","alias_value":"XZDXCK6WABNO","created_at":"2026-06-08T01:03:48.814279+00:00"},{"alias_kind":"pith_short_16","alias_value":"XZDXCK6WABNOITK5","created_at":"2026-06-08T01:03:48.814279+00:00"},{"alias_kind":"pith_short_8","alias_value":"XZDXCK6W","created_at":"2026-06-08T01:03:48.814279+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2605.13612","citing_title":"Deep Learning as Neural Low-Degree Filtering: A Spectral Theory of Hierarchical Feature Learning","ref_index":78,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06826","citing_title":"How Does Attention Help? Insights from Random Matrices on Signal Recovery from Sequence Models","ref_index":1,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XZDXCK6WABNOITK5ZSVKRRDCN2","json":"https://pith.science/pith/XZDXCK6WABNOITK5ZSVKRRDCN2.json","graph_json":"https://pith.science/api/pith-number/XZDXCK6WABNOITK5ZSVKRRDCN2/graph.json","events_json":"https://pith.science/api/pith-number/XZDXCK6WABNOITK5ZSVKRRDCN2/events.json","paper":"https://pith.science/paper/XZDXCK6W"},"agent_actions":{"view_html":"https://pith.science/pith/XZDXCK6WABNOITK5ZSVKRRDCN2","download_json":"https://pith.science/pith/XZDXCK6WABNOITK5ZSVKRRDCN2.json","view_paper":"https://pith.science/paper/XZDXCK6W","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.24914&json=true","fetch_graph":"https://pith.science/api/pith-number/XZDXCK6WABNOITK5ZSVKRRDCN2/graph.json","fetch_events":"https://pith.science/api/pith-number/XZDXCK6WABNOITK5ZSVKRRDCN2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XZDXCK6WABNOITK5ZSVKRRDCN2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XZDXCK6WABNOITK5ZSVKRRDCN2/action/storage_attestation","attest_author":"https://pith.science/pith/XZDXCK6WABNOITK5ZSVKRRDCN2/action/author_attestation","sign_citation":"https://pith.science/pith/XZDXCK6WABNOITK5ZSVKRRDCN2/action/citation_signature","submit_replication":"https://pith.science/pith/XZDXCK6WABNOITK5ZSVKRRDCN2/action/replication_record"}},"created_at":"2026-06-08T01:03:48.814279+00:00","updated_at":"2026-06-08T01:03:48.814279+00:00"}