{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:D2AMNJKCX2ARAK2UGW4Y37JCKJ","short_pith_number":"pith:D2AMNJKC","schema_version":"1.0","canonical_sha256":"1e80c6a542be81102b5435b98dfd225241e9ea5e1f9dc926a13ecd7925592b17","source":{"kind":"arxiv","id":"2510.01718","version":2},"attestation_state":"computed","paper":{"title":"Accelerating Attention with Basis Decomposition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Jialin Zhao","submitted_at":"2025-10-02T06:58:10Z","abstract_excerpt":"Attention is a core operation in large language models (LLMs). We present BD Attention (BDA), a lossless algorithmic reformulation of attention. BDA is enabled by a simple matrix identity from Basis Decomposition (BD), which restructures multi-head projections into a compact form while preserving exact outputs. Unlike I/O-aware system optimizations such as FlashAttention, BDA provides a mathematically guaranteed acceleration that is architecture-agnostic. On DeepSeek-V2-Lite (16B, FP16), BDA requires only 4s of offline preparation with no retraining required and, on modern GPUs, achieves 34% f"},"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":"2510.01718","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-10-02T06:58:10Z","cross_cats_sorted":[],"title_canon_sha256":"0eca7f6695d2c089466c3251a8543076811445623ca8e2fb4a28ee004ff98569","abstract_canon_sha256":"ed8493d0c25c5d2e581e7bea82b4c3c6aee4bfe778cfd9301100e437b65bd6e3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-29T01:14:24.534517Z","signature_b64":"BU1h4XW0ga4ijm7VIJwfXkiYFOOpKgsLJaBMOV9FEURbOfkkrg5vlqD8IJiY0a+VUxOLwMNBVhRr1XRhuhsaCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1e80c6a542be81102b5435b98dfd225241e9ea5e1f9dc926a13ecd7925592b17","last_reissued_at":"2026-06-29T01:14:24.534058Z","signature_status":"signed_v1","first_computed_at":"2026-06-29T01:14:24.534058Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Accelerating Attention with Basis Decomposition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Jialin Zhao","submitted_at":"2025-10-02T06:58:10Z","abstract_excerpt":"Attention is a core operation in large language models (LLMs). We present BD Attention (BDA), a lossless algorithmic reformulation of attention. BDA is enabled by a simple matrix identity from Basis Decomposition (BD), which restructures multi-head projections into a compact form while preserving exact outputs. Unlike I/O-aware system optimizations such as FlashAttention, BDA provides a mathematically guaranteed acceleration that is architecture-agnostic. On DeepSeek-V2-Lite (16B, FP16), BDA requires only 4s of offline preparation with no retraining required and, on modern GPUs, achieves 34% f"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2510.01718","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/2510.01718/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":"2510.01718","created_at":"2026-06-29T01:14:24.534117+00:00"},{"alias_kind":"arxiv_version","alias_value":"2510.01718v2","created_at":"2026-06-29T01:14:24.534117+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2510.01718","created_at":"2026-06-29T01:14:24.534117+00:00"},{"alias_kind":"pith_short_12","alias_value":"D2AMNJKCX2AR","created_at":"2026-06-29T01:14:24.534117+00:00"},{"alias_kind":"pith_short_16","alias_value":"D2AMNJKCX2ARAK2U","created_at":"2026-06-29T01:14:24.534117+00:00"},{"alias_kind":"pith_short_8","alias_value":"D2AMNJKC","created_at":"2026-06-29T01:14:24.534117+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/D2AMNJKCX2ARAK2UGW4Y37JCKJ","json":"https://pith.science/pith/D2AMNJKCX2ARAK2UGW4Y37JCKJ.json","graph_json":"https://pith.science/api/pith-number/D2AMNJKCX2ARAK2UGW4Y37JCKJ/graph.json","events_json":"https://pith.science/api/pith-number/D2AMNJKCX2ARAK2UGW4Y37JCKJ/events.json","paper":"https://pith.science/paper/D2AMNJKC"},"agent_actions":{"view_html":"https://pith.science/pith/D2AMNJKCX2ARAK2UGW4Y37JCKJ","download_json":"https://pith.science/pith/D2AMNJKCX2ARAK2UGW4Y37JCKJ.json","view_paper":"https://pith.science/paper/D2AMNJKC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2510.01718&json=true","fetch_graph":"https://pith.science/api/pith-number/D2AMNJKCX2ARAK2UGW4Y37JCKJ/graph.json","fetch_events":"https://pith.science/api/pith-number/D2AMNJKCX2ARAK2UGW4Y37JCKJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D2AMNJKCX2ARAK2UGW4Y37JCKJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D2AMNJKCX2ARAK2UGW4Y37JCKJ/action/storage_attestation","attest_author":"https://pith.science/pith/D2AMNJKCX2ARAK2UGW4Y37JCKJ/action/author_attestation","sign_citation":"https://pith.science/pith/D2AMNJKCX2ARAK2UGW4Y37JCKJ/action/citation_signature","submit_replication":"https://pith.science/pith/D2AMNJKCX2ARAK2UGW4Y37JCKJ/action/replication_record"}},"created_at":"2026-06-29T01:14:24.534117+00:00","updated_at":"2026-06-29T01:14:24.534117+00:00"}