{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:T3IGRLS3BGJ3CVCX4BO7CUJP7D","short_pith_number":"pith:T3IGRLS3","schema_version":"1.0","canonical_sha256":"9ed068ae5b0993b15457e05df1512ff8ffd0b5fc6524c68cd8260997dcd467f9","source":{"kind":"arxiv","id":"2510.05554","version":2},"attestation_state":"computed","paper":{"title":"Critical attention scaling in long-context transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.DM","math.CA"],"primary_cat":"cs.LG","authors_text":"Philippe Rigollet, Shi Chen, Yury Polyanskiy, Zhengjiang Lin","submitted_at":"2025-10-07T03:51:57Z","abstract_excerpt":"As large language models scale to longer contexts, attention layers suffer from a fundamental pathology: attention scores collapse toward uniformity as context length $n$ increases, causing tokens to cluster excessively, a phenomenon known as rank-collapse. While $\\textit{attention scaling}$ effectively addresses this deficiency by rescaling attention scores with a polylogarithmic factor $\\beta_n$, theoretical justification for this approach remains lacking.\n  We analyze a simplified yet tractable model that magnifies the effect of attention scaling. In this model, attention exhibits a phase t"},"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":"2510.05554","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-10-07T03:51:57Z","cross_cats_sorted":["cs.AI","cs.DM","math.CA"],"title_canon_sha256":"e8597fec481ada6913ed147f94420ebe3dab9bf165da97c0524cf36d9a95d181","abstract_canon_sha256":"9940b335e755a44805f4225e15c59d6b655e967b00374834b195355afc1e73e8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9ed068ae5b0993b15457e05df1512ff8ffd0b5fc6524c68cd8260997dcd467f9","last_reissued_at":"2026-07-31T01:08:51.370122Z","signature_status":"unsigned_v0","first_computed_at":"2026-07-31T01:08:51.370122Z"},"graph_snapshot":{"paper":{"title":"Critical attention scaling in long-context transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.DM","math.CA"],"primary_cat":"cs.LG","authors_text":"Philippe Rigollet, Shi Chen, Yury Polyanskiy, Zhengjiang Lin","submitted_at":"2025-10-07T03:51:57Z","abstract_excerpt":"As large language models scale to longer contexts, attention layers suffer from a fundamental pathology: attention scores collapse toward uniformity as context length $n$ increases, causing tokens to cluster excessively, a phenomenon known as rank-collapse. While $\\textit{attention scaling}$ effectively addresses this deficiency by rescaling attention scores with a polylogarithmic factor $\\beta_n$, theoretical justification for this approach remains lacking.\n  We analyze a simplified yet tractable model that magnifies the effect of attention scaling. In this model, attention exhibits a phase t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.05554","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/2510.05554/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":"2510.05554","created_at":"2026-07-31T01:08:51.373187+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.05554v2","created_at":"2026-07-31T01:08:51.373187+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.05554","created_at":"2026-07-31T01:08:51.373187+00:00"},{"alias_kind":"pith_short_12","alias_value":"T3IGRLS3BGJ3","created_at":"2026-07-31T01:08:51.373187+00:00"},{"alias_kind":"pith_short_16","alias_value":"T3IGRLS3BGJ3CVCX","created_at":"2026-07-31T01:08:51.373187+00:00"},{"alias_kind":"pith_short_8","alias_value":"T3IGRLS3","created_at":"2026-07-31T01:08:51.373187+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"2601.21366","citing_title":"Perceptrons and localization of attention's mean-field landscape","ref_index":6,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12697","citing_title":"A Unified Framework for Critical Scaling of Inverse Temperature in Self-Attention","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2605.10931","citing_title":"Quantifying Concentration Phenomena of Mean-Field Transformers in the Low-Temperature Regime","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2605.06509","citing_title":"FreeSpec: Training-Free Long Video Generation via Singular-Spectrum Reconstruction","ref_index":23,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T3IGRLS3BGJ3CVCX4BO7CUJP7D","json":"https://pith.science/pith/T3IGRLS3BGJ3CVCX4BO7CUJP7D.json","graph_json":"https://pith.science/api/pith-number/T3IGRLS3BGJ3CVCX4BO7CUJP7D/graph.json","events_json":"https://pith.science/api/pith-number/T3IGRLS3BGJ3CVCX4BO7CUJP7D/events.json","paper":"https://pith.science/paper/T3IGRLS3"},"agent_actions":{"view_html":"https://pith.science/pith/T3IGRLS3BGJ3CVCX4BO7CUJP7D","download_json":"https://pith.science/pith/T3IGRLS3BGJ3CVCX4BO7CUJP7D.json","view_paper":"https://pith.science/paper/T3IGRLS3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.05554&json=true","fetch_graph":"https://pith.science/api/pith-number/T3IGRLS3BGJ3CVCX4BO7CUJP7D/graph.json","fetch_events":"https://pith.science/api/pith-number/T3IGRLS3BGJ3CVCX4BO7CUJP7D/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T3IGRLS3BGJ3CVCX4BO7CUJP7D/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T3IGRLS3BGJ3CVCX4BO7CUJP7D/action/storage_attestation","attest_author":"https://pith.science/pith/T3IGRLS3BGJ3CVCX4BO7CUJP7D/action/author_attestation","sign_citation":"https://pith.science/pith/T3IGRLS3BGJ3CVCX4BO7CUJP7D/action/citation_signature","submit_replication":"https://pith.science/pith/T3IGRLS3BGJ3CVCX4BO7CUJP7D/action/replication_record"}},"created_at":"2026-07-31T01:08:51.373187+00:00","updated_at":"2026-07-31T01:08:51.373187+00:00"}