{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:54D5U3DIHHX53DABESULOSJXFB","short_pith_number":"pith:54D5U3DI","schema_version":"1.0","canonical_sha256":"ef07da6c6839efdd8c0124a8b74937284dd62a2244fb5e2bb7648b4f94abd9a1","source":{"kind":"arxiv","id":"2512.10656","version":3},"attestation_state":"computed","paper":{"title":"Token Sample Complexity of Attention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Cyril Letrouit, Fran\\c{c}ois-Xavier Vialard, Gabriel Peyr\\'e, L\\'ea Bohbot","submitted_at":"2025-12-11T14:02:34Z","abstract_excerpt":"As context windows in large language models continue to expand, it is essential to characterize how attention behaves at extreme sequence lengths. We introduce token sample complexity: the rate at which attention computed on $n$ tokens converges to its infinite-token limit. We estimate finite-$n$ convergence bounds at two levels: pointwise uniform convergence of the attention map, and convergence of moments for the transformed token distribution.\n  For compactly supported (and more generally sub-Gaussian) distributions, our first result shows that the attention map converges uniformly on a bal"},"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":"2512.10656","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-12-11T14:02:34Z","cross_cats_sorted":[],"title_canon_sha256":"f29d96545ebfce6096a898f47a7ebf76b67aa005be6f2a183e4382ed9808c26c","abstract_canon_sha256":"7616453bf1047b94d5a09b6b1b2e696d7e7612dca7e2f75a06a33e14b5b818ac"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T02:07:15.120925Z","signature_b64":"ugMWwSF0SJmaUT8mUxmIf1Qi+LMl6YmDJ2le1w3cFy4YxQbls1qGsndvWzIDjQ9isEWXt68uR0pokB3jILXZCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ef07da6c6839efdd8c0124a8b74937284dd62a2244fb5e2bb7648b4f94abd9a1","last_reissued_at":"2026-06-09T02:07:15.119825Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T02:07:15.119825Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Token Sample Complexity of Attention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Cyril Letrouit, Fran\\c{c}ois-Xavier Vialard, Gabriel Peyr\\'e, L\\'ea Bohbot","submitted_at":"2025-12-11T14:02:34Z","abstract_excerpt":"As context windows in large language models continue to expand, it is essential to characterize how attention behaves at extreme sequence lengths. We introduce token sample complexity: the rate at which attention computed on $n$ tokens converges to its infinite-token limit. We estimate finite-$n$ convergence bounds at two levels: pointwise uniform convergence of the attention map, and convergence of moments for the transformed token distribution.\n  For compactly supported (and more generally sub-Gaussian) distributions, our first result shows that the attention map converges uniformly on a bal"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2512.10656","kind":"arxiv","version":3},"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/2512.10656/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":"2512.10656","created_at":"2026-06-09T02:07:15.119982+00:00"},{"alias_kind":"arxiv_version","alias_value":"2512.10656v3","created_at":"2026-06-09T02:07:15.119982+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2512.10656","created_at":"2026-06-09T02:07:15.119982+00:00"},{"alias_kind":"pith_short_12","alias_value":"54D5U3DIHHX5","created_at":"2026-06-09T02:07:15.119982+00:00"},{"alias_kind":"pith_short_16","alias_value":"54D5U3DIHHX53DAB","created_at":"2026-06-09T02:07:15.119982+00:00"},{"alias_kind":"pith_short_8","alias_value":"54D5U3DI","created_at":"2026-06-09T02:07:15.119982+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.16747","citing_title":"Propagation of Chaos in Contextual Flow Maps","ref_index":5,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/54D5U3DIHHX53DABESULOSJXFB","json":"https://pith.science/pith/54D5U3DIHHX53DABESULOSJXFB.json","graph_json":"https://pith.science/api/pith-number/54D5U3DIHHX53DABESULOSJXFB/graph.json","events_json":"https://pith.science/api/pith-number/54D5U3DIHHX53DABESULOSJXFB/events.json","paper":"https://pith.science/paper/54D5U3DI"},"agent_actions":{"view_html":"https://pith.science/pith/54D5U3DIHHX53DABESULOSJXFB","download_json":"https://pith.science/pith/54D5U3DIHHX53DABESULOSJXFB.json","view_paper":"https://pith.science/paper/54D5U3DI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2512.10656&json=true","fetch_graph":"https://pith.science/api/pith-number/54D5U3DIHHX53DABESULOSJXFB/graph.json","fetch_events":"https://pith.science/api/pith-number/54D5U3DIHHX53DABESULOSJXFB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/54D5U3DIHHX53DABESULOSJXFB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/54D5U3DIHHX53DABESULOSJXFB/action/storage_attestation","attest_author":"https://pith.science/pith/54D5U3DIHHX53DABESULOSJXFB/action/author_attestation","sign_citation":"https://pith.science/pith/54D5U3DIHHX53DABESULOSJXFB/action/citation_signature","submit_replication":"https://pith.science/pith/54D5U3DIHHX53DABESULOSJXFB/action/replication_record"}},"created_at":"2026-06-09T02:07:15.119982+00:00","updated_at":"2026-06-09T02:07:15.119982+00:00"}