{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:K4FBSSCUWVPHP3P7L26EKDVBXW","short_pith_number":"pith:K4FBSSCU","schema_version":"1.0","canonical_sha256":"570a194854b55e77edff5ebc450ea1bda837c3a09514dc8e65a388a1ba3b2b8b","source":{"kind":"arxiv","id":"2202.10447","version":2},"attestation_state":"computed","paper":{"title":"Transformer Quality in Linear Time","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.NE"],"primary_cat":"cs.LG","authors_text":"Hanxiao Liu, Quoc V. Le, Weizhe Hua, Zihang Dai","submitted_at":"2022-02-21T18:59:38Z","abstract_excerpt":"We revisit the design choices in Transformers, and propose methods to address their weaknesses in handling long sequences. First, we propose a simple layer named gated attention unit, which allows the use of a weaker single-head attention with minimal quality loss. We then propose a linear approximation method complementary to this new layer, which is accelerator-friendly and highly competitive in quality. The resulting model, named FLASH, matches the perplexity of improved Transformers over both short (512) and long (8K) context lengths, achieving training speedups of up to 4.9$\\times$ on Wik"},"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":"2202.10447","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2022-02-21T18:59:38Z","cross_cats_sorted":["cs.AI","cs.CL","cs.NE"],"title_canon_sha256":"021ddd5d093ce5e982bf4c58935925ec42a9039bc10bbef2aaf967f332d77c6d","abstract_canon_sha256":"f211b3640d8a6674e3a59e0ecebb9d8a45552371e5ffb01e56a9228503c2fa83"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:34:50.875246Z","signature_b64":"Lbyly/4AH+ZltGTiL31YUhRbCUbAk5Fz1b4CTF1z34AiOJPxMEP+hr1N5QYJrUyEL+ddux/exJjkFf4vY9YKDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"570a194854b55e77edff5ebc450ea1bda837c3a09514dc8e65a388a1ba3b2b8b","last_reissued_at":"2026-07-05T04:34:50.874755Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:34:50.874755Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Transformer Quality in Linear Time","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.NE"],"primary_cat":"cs.LG","authors_text":"Hanxiao Liu, Quoc V. Le, Weizhe Hua, Zihang Dai","submitted_at":"2022-02-21T18:59:38Z","abstract_excerpt":"We revisit the design choices in Transformers, and propose methods to address their weaknesses in handling long sequences. First, we propose a simple layer named gated attention unit, which allows the use of a weaker single-head attention with minimal quality loss. We then propose a linear approximation method complementary to this new layer, which is accelerator-friendly and highly competitive in quality. The resulting model, named FLASH, matches the perplexity of improved Transformers over both short (512) and long (8K) context lengths, achieving training speedups of up to 4.9$\\times$ on Wik"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2202.10447","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/2202.10447/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":"2202.10447","created_at":"2026-07-05T04:34:50.874817+00:00"},{"alias_kind":"arxiv_version","alias_value":"2202.10447v2","created_at":"2026-07-05T04:34:50.874817+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2202.10447","created_at":"2026-07-05T04:34:50.874817+00:00"},{"alias_kind":"pith_short_12","alias_value":"K4FBSSCUWVPH","created_at":"2026-07-05T04:34:50.874817+00:00"},{"alias_kind":"pith_short_16","alias_value":"K4FBSSCUWVPHP3P7","created_at":"2026-07-05T04:34:50.874817+00:00"},{"alias_kind":"pith_short_8","alias_value":"K4FBSSCU","created_at":"2026-07-05T04:34:50.874817+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2509.22630","citing_title":"StateX: Enhancing RNN Recall via Post-training State Expansion","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2205.14135","citing_title":"FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18580","citing_title":"Sessa: Selective State Space Attention","ref_index":31,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/K4FBSSCUWVPHP3P7L26EKDVBXW","json":"https://pith.science/pith/K4FBSSCUWVPHP3P7L26EKDVBXW.json","graph_json":"https://pith.science/api/pith-number/K4FBSSCUWVPHP3P7L26EKDVBXW/graph.json","events_json":"https://pith.science/api/pith-number/K4FBSSCUWVPHP3P7L26EKDVBXW/events.json","paper":"https://pith.science/paper/K4FBSSCU"},"agent_actions":{"view_html":"https://pith.science/pith/K4FBSSCUWVPHP3P7L26EKDVBXW","download_json":"https://pith.science/pith/K4FBSSCUWVPHP3P7L26EKDVBXW.json","view_paper":"https://pith.science/paper/K4FBSSCU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2202.10447&json=true","fetch_graph":"https://pith.science/api/pith-number/K4FBSSCUWVPHP3P7L26EKDVBXW/graph.json","fetch_events":"https://pith.science/api/pith-number/K4FBSSCUWVPHP3P7L26EKDVBXW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/K4FBSSCUWVPHP3P7L26EKDVBXW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/K4FBSSCUWVPHP3P7L26EKDVBXW/action/storage_attestation","attest_author":"https://pith.science/pith/K4FBSSCUWVPHP3P7L26EKDVBXW/action/author_attestation","sign_citation":"https://pith.science/pith/K4FBSSCUWVPHP3P7L26EKDVBXW/action/citation_signature","submit_replication":"https://pith.science/pith/K4FBSSCUWVPHP3P7L26EKDVBXW/action/replication_record"}},"created_at":"2026-07-05T04:34:50.874817+00:00","updated_at":"2026-07-05T04:34:50.874817+00:00"}