{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:AVE6KDIV2OPOG473BH6ANCUCYE","short_pith_number":"pith:AVE6KDIV","schema_version":"1.0","canonical_sha256":"0549e50d15d39ee373fb09fc068a82c12fcbcb5cbf8aea375b701a276b4f2ec2","source":{"kind":"arxiv","id":"2405.05219","version":2},"attestation_state":"computed","paper":{"title":"Conv-Basis: A New Paradigm for Efficient Attention Inference and Gradient Computation in Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Heshan Liu, Junze Yin, Yingyu Liang, Zhao Song, Zhenmei Shi, Zhuoyan Xu","submitted_at":"2024-05-08T17:11:38Z","abstract_excerpt":"The self-attention mechanism is the key to the success of transformers in recent Large Language Models (LLMs). However, the quadratic computational cost $O(n^2)$ in the input sequence length $n$ is a notorious obstacle for further improvement and scalability in longer contexts. In this work, we leverage the convolution-like structure of attention matrices to develop an efficient approximation method for attention computation using convolution matrices. We propose a $\\mathsf{conv}$ basis system, analogous to the rank basis, and show that any lower triangular matrix can always be decomposed as a"},"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":"2405.05219","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-05-08T17:11:38Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"eacf7891e097a7c9a48a1d08394b08bf6006aaaa3c7a8f07f52366a75b1b05a6","abstract_canon_sha256":"e25b3853b7c283ee7e5e640e96f8198a3ef966ca9e605dca3b8baf28f4afaccd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:21:04.249429Z","signature_b64":"YiDhv3Qbhcd9dF85iaont+YTLNUJQKEP7no1jCAz0mE6QWiiEnRTSx5GEjclqRpuGijCF/gXm6j9SgLmqOw/Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0549e50d15d39ee373fb09fc068a82c12fcbcb5cbf8aea375b701a276b4f2ec2","last_reissued_at":"2026-07-05T09:21:04.248926Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:21:04.248926Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Conv-Basis: A New Paradigm for Efficient Attention Inference and Gradient Computation in Transformers","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Heshan Liu, Junze Yin, Yingyu Liang, Zhao Song, Zhenmei Shi, Zhuoyan Xu","submitted_at":"2024-05-08T17:11:38Z","abstract_excerpt":"The self-attention mechanism is the key to the success of transformers in recent Large Language Models (LLMs). However, the quadratic computational cost $O(n^2)$ in the input sequence length $n$ is a notorious obstacle for further improvement and scalability in longer contexts. In this work, we leverage the convolution-like structure of attention matrices to develop an efficient approximation method for attention computation using convolution matrices. We propose a $\\mathsf{conv}$ basis system, analogous to the rank basis, and show that any lower triangular matrix can always be decomposed as a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.05219","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/2405.05219/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":"2405.05219","created_at":"2026-07-05T09:21:04.248990+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.05219v2","created_at":"2026-07-05T09:21:04.248990+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.05219","created_at":"2026-07-05T09:21:04.248990+00:00"},{"alias_kind":"pith_short_12","alias_value":"AVE6KDIV2OPO","created_at":"2026-07-05T09:21:04.248990+00:00"},{"alias_kind":"pith_short_16","alias_value":"AVE6KDIV2OPOG473","created_at":"2026-07-05T09:21:04.248990+00:00"},{"alias_kind":"pith_short_8","alias_value":"AVE6KDIV","created_at":"2026-07-05T09:21:04.248990+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.19993","citing_title":"CoVE: Compressed Vocabulary Expansion Makes Better LLM-based Recommender Systems","ref_index":54,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AVE6KDIV2OPOG473BH6ANCUCYE","json":"https://pith.science/pith/AVE6KDIV2OPOG473BH6ANCUCYE.json","graph_json":"https://pith.science/api/pith-number/AVE6KDIV2OPOG473BH6ANCUCYE/graph.json","events_json":"https://pith.science/api/pith-number/AVE6KDIV2OPOG473BH6ANCUCYE/events.json","paper":"https://pith.science/paper/AVE6KDIV"},"agent_actions":{"view_html":"https://pith.science/pith/AVE6KDIV2OPOG473BH6ANCUCYE","download_json":"https://pith.science/pith/AVE6KDIV2OPOG473BH6ANCUCYE.json","view_paper":"https://pith.science/paper/AVE6KDIV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.05219&json=true","fetch_graph":"https://pith.science/api/pith-number/AVE6KDIV2OPOG473BH6ANCUCYE/graph.json","fetch_events":"https://pith.science/api/pith-number/AVE6KDIV2OPOG473BH6ANCUCYE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AVE6KDIV2OPOG473BH6ANCUCYE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AVE6KDIV2OPOG473BH6ANCUCYE/action/storage_attestation","attest_author":"https://pith.science/pith/AVE6KDIV2OPOG473BH6ANCUCYE/action/author_attestation","sign_citation":"https://pith.science/pith/AVE6KDIV2OPOG473BH6ANCUCYE/action/citation_signature","submit_replication":"https://pith.science/pith/AVE6KDIV2OPOG473BH6ANCUCYE/action/replication_record"}},"created_at":"2026-07-05T09:21:04.248990+00:00","updated_at":"2026-07-05T09:21:04.248990+00:00"}