{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:LTYDHCQP44HDTJGTYZGXF2CPQ4","short_pith_number":"pith:LTYDHCQP","schema_version":"1.0","canonical_sha256":"5cf0338a0fe70e39a4d3c64d72e84f872c2e58d1cfcbd12f12662095bea7568a","source":{"kind":"arxiv","id":"2509.10406","version":4},"attestation_state":"computed","paper":{"title":"Multipole Semantic Attention: A Fast Approximation of Softmax Attention for Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Kristian Kersting, Rupert Mitchell","submitted_at":"2025-09-12T16:58:17Z","abstract_excerpt":"Pretraining transformers on long sequences (entire code repositories, collections of related documents) is bottlenecked by quadratic attention costs. We present Multipole Semantic Attention (MuSe), which accelerates 64k-context pretraining by 36% while matching baseline loss, requiring no architectural changes. MuSe clusters queries and keys separately in representation space. This yields query-specific summaries that substantially outperform spatial blocking at matched sparsity, while also enabling drop-in compatibility with existing pretrained models; we validate on Llama 3.1-8B and 3.2-1B w"},"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":"2509.10406","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-09-12T16:58:17Z","cross_cats_sorted":[],"title_canon_sha256":"bcafed616852da6d7360fdb40c8078f40f0c7e9d484be5c03dc00a6fc807c977","abstract_canon_sha256":"f2d3063b0a5f329b6a8917d936528b4bca2066cb195dbcdcd9c3acbd502bb5ac"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-01T01:17:41.358954Z","signature_b64":"lACleOOKqzTwOC+3h1sIN8zC2pEKgzciKr+vRj0Paj7rzizjtoutNq33LvoekjKfutsd523wW2PrVqfm/I69Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5cf0338a0fe70e39a4d3c64d72e84f872c2e58d1cfcbd12f12662095bea7568a","last_reissued_at":"2026-07-01T01:17:41.358451Z","signature_status":"signed_v1","first_computed_at":"2026-07-01T01:17:41.358451Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multipole Semantic Attention: A Fast Approximation of Softmax Attention for Pretraining","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Kristian Kersting, Rupert Mitchell","submitted_at":"2025-09-12T16:58:17Z","abstract_excerpt":"Pretraining transformers on long sequences (entire code repositories, collections of related documents) is bottlenecked by quadratic attention costs. We present Multipole Semantic Attention (MuSe), which accelerates 64k-context pretraining by 36% while matching baseline loss, requiring no architectural changes. MuSe clusters queries and keys separately in representation space. This yields query-specific summaries that substantially outperform spatial blocking at matched sparsity, while also enabling drop-in compatibility with existing pretrained models; we validate on Llama 3.1-8B and 3.2-1B w"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.10406","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/2509.10406/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":"2509.10406","created_at":"2026-07-01T01:17:41.358509+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.10406v4","created_at":"2026-07-01T01:17:41.358509+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.10406","created_at":"2026-07-01T01:17:41.358509+00:00"},{"alias_kind":"pith_short_12","alias_value":"LTYDHCQP44HD","created_at":"2026-07-01T01:17:41.358509+00:00"},{"alias_kind":"pith_short_16","alias_value":"LTYDHCQP44HDTJGT","created_at":"2026-07-01T01:17:41.358509+00:00"},{"alias_kind":"pith_short_8","alias_value":"LTYDHCQP","created_at":"2026-07-01T01:17:41.358509+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LTYDHCQP44HDTJGTYZGXF2CPQ4","json":"https://pith.science/pith/LTYDHCQP44HDTJGTYZGXF2CPQ4.json","graph_json":"https://pith.science/api/pith-number/LTYDHCQP44HDTJGTYZGXF2CPQ4/graph.json","events_json":"https://pith.science/api/pith-number/LTYDHCQP44HDTJGTYZGXF2CPQ4/events.json","paper":"https://pith.science/paper/LTYDHCQP"},"agent_actions":{"view_html":"https://pith.science/pith/LTYDHCQP44HDTJGTYZGXF2CPQ4","download_json":"https://pith.science/pith/LTYDHCQP44HDTJGTYZGXF2CPQ4.json","view_paper":"https://pith.science/paper/LTYDHCQP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.10406&json=true","fetch_graph":"https://pith.science/api/pith-number/LTYDHCQP44HDTJGTYZGXF2CPQ4/graph.json","fetch_events":"https://pith.science/api/pith-number/LTYDHCQP44HDTJGTYZGXF2CPQ4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LTYDHCQP44HDTJGTYZGXF2CPQ4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LTYDHCQP44HDTJGTYZGXF2CPQ4/action/storage_attestation","attest_author":"https://pith.science/pith/LTYDHCQP44HDTJGTYZGXF2CPQ4/action/author_attestation","sign_citation":"https://pith.science/pith/LTYDHCQP44HDTJGTYZGXF2CPQ4/action/citation_signature","submit_replication":"https://pith.science/pith/LTYDHCQP44HDTJGTYZGXF2CPQ4/action/replication_record"}},"created_at":"2026-07-01T01:17:41.358509+00:00","updated_at":"2026-07-01T01:17:41.358509+00:00"}