{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:OTPPFXCN7IZAJS3HP2K2RTLUG3","short_pith_number":"pith:OTPPFXCN","schema_version":"1.0","canonical_sha256":"74def2dc4dfa3204cb677e95a8cd7436f5a076d6cca395aac87aa60a92f3132b","source":{"kind":"arxiv","id":"2507.04239","version":1},"attestation_state":"computed","paper":{"title":"Scaling Context Requires Rethinking Attention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Carles Gelada, Jacob Buckman, Sean Zhang, Txus Bach","submitted_at":"2025-07-06T04:15:34Z","abstract_excerpt":"We argue that neither transformers nor sub-quadratic architectures are well suited to training at long sequence lengths: the cost of processing the context is too expensive in the former, too inexpensive in the latter. Approaches such as sliding window attention which reduce the cost-per-token of a transformer impair in-context learning, and so are also unsuitable. To address these limitations, we introduce power attention, an architectural layer for linear-cost sequence modeling whose state size can be adjusted independently of parameters, unlocking the advantages of linear attention on pract"},"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":"2507.04239","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-07-06T04:15:34Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"d23e91819c7d702d0dfcce21732292b8b589055a68ba8b4bc0203d56b12d0ecb","abstract_canon_sha256":"1276c1f8ee0f10f7ece6d6c8e51376033e66782eecab7cfec8c18e0d673653db"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:32:44.060165Z","signature_b64":"KSpCfIx2QplRQZ4bx9/N/BKixnrEXK1p6zhOcGPSqfqbajyr4kblFXNdTxebIPdPM81LdCKaT3lQRSnF3QY4Bw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"74def2dc4dfa3204cb677e95a8cd7436f5a076d6cca395aac87aa60a92f3132b","last_reissued_at":"2026-07-05T11:32:44.059714Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:32:44.059714Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Context Requires Rethinking Attention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Carles Gelada, Jacob Buckman, Sean Zhang, Txus Bach","submitted_at":"2025-07-06T04:15:34Z","abstract_excerpt":"We argue that neither transformers nor sub-quadratic architectures are well suited to training at long sequence lengths: the cost of processing the context is too expensive in the former, too inexpensive in the latter. Approaches such as sliding window attention which reduce the cost-per-token of a transformer impair in-context learning, and so are also unsuitable. To address these limitations, we introduce power attention, an architectural layer for linear-cost sequence modeling whose state size can be adjusted independently of parameters, unlocking the advantages of linear attention on pract"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.04239","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/2507.04239/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":"2507.04239","created_at":"2026-07-05T11:32:44.059766+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.04239v1","created_at":"2026-07-05T11:32:44.059766+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.04239","created_at":"2026-07-05T11:32:44.059766+00:00"},{"alias_kind":"pith_short_12","alias_value":"OTPPFXCN7IZA","created_at":"2026-07-05T11:32:44.059766+00:00"},{"alias_kind":"pith_short_16","alias_value":"OTPPFXCN7IZAJS3H","created_at":"2026-07-05T11:32:44.059766+00:00"},{"alias_kind":"pith_short_8","alias_value":"OTPPFXCN","created_at":"2026-07-05T11:32:44.059766+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.15156","citing_title":"MeMo: Memory as a Model","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15156","citing_title":"MeMo: Memory as a Model","ref_index":40,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/OTPPFXCN7IZAJS3HP2K2RTLUG3","json":"https://pith.science/pith/OTPPFXCN7IZAJS3HP2K2RTLUG3.json","graph_json":"https://pith.science/api/pith-number/OTPPFXCN7IZAJS3HP2K2RTLUG3/graph.json","events_json":"https://pith.science/api/pith-number/OTPPFXCN7IZAJS3HP2K2RTLUG3/events.json","paper":"https://pith.science/paper/OTPPFXCN"},"agent_actions":{"view_html":"https://pith.science/pith/OTPPFXCN7IZAJS3HP2K2RTLUG3","download_json":"https://pith.science/pith/OTPPFXCN7IZAJS3HP2K2RTLUG3.json","view_paper":"https://pith.science/paper/OTPPFXCN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.04239&json=true","fetch_graph":"https://pith.science/api/pith-number/OTPPFXCN7IZAJS3HP2K2RTLUG3/graph.json","fetch_events":"https://pith.science/api/pith-number/OTPPFXCN7IZAJS3HP2K2RTLUG3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/OTPPFXCN7IZAJS3HP2K2RTLUG3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/OTPPFXCN7IZAJS3HP2K2RTLUG3/action/storage_attestation","attest_author":"https://pith.science/pith/OTPPFXCN7IZAJS3HP2K2RTLUG3/action/author_attestation","sign_citation":"https://pith.science/pith/OTPPFXCN7IZAJS3HP2K2RTLUG3/action/citation_signature","submit_replication":"https://pith.science/pith/OTPPFXCN7IZAJS3HP2K2RTLUG3/action/replication_record"}},"created_at":"2026-07-05T11:32:44.059766+00:00","updated_at":"2026-07-05T11:32:44.059766+00:00"}