{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:ZPDLBDPHLR45TNZWU5YF4JDNZV","short_pith_number":"pith:ZPDLBDPH","schema_version":"1.0","canonical_sha256":"cbc6b08de75c79d9b736a7705e246dcd7bdc967d75c683396abb4b813a68fafe","source":{"kind":"arxiv","id":"2107.05768","version":2},"attestation_state":"computed","paper":{"title":"Combiner: Full Attention Transformer with Sparse Computation Cost","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Bo Dai, Dale Schuurmans, Hanjun Dai, Hongyu Ren, Jure Leskovec, Mengjiao Yang, Zihang Dai","submitted_at":"2021-07-12T22:43:11Z","abstract_excerpt":"Transformers provide a class of expressive architectures that are extremely effective for sequence modeling. However, the key limitation of transformers is their quadratic memory and time complexity $\\mathcal{O}(L^2)$ with respect to the sequence length in attention layers, which restricts application in extremely long sequences. Most existing approaches leverage sparsity or low-rank assumptions in the attention matrix to reduce cost, but sacrifice expressiveness. Instead, we propose Combiner, which provides full attention capability in each attention head while maintaining low computation and"},"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":"2107.05768","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-07-12T22:43:11Z","cross_cats_sorted":["cs.CL","cs.CV"],"title_canon_sha256":"ebac4669d143c189c653453783a43c350ff26db8653524bb06a3e00786f66f2e","abstract_canon_sha256":"fc94c228fd5192010b461be332fc29295c8e0ec858a659264c52fdf329375b49"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:26:42.754454Z","signature_b64":"v8Q90EpBh0LoXO1rKgNT7m1kDuLg8gKhKJRAPa9F9SjRqlQt1yf/tMK3LE39RFInyWP2UUg1n2iTDlblgXt8Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cbc6b08de75c79d9b736a7705e246dcd7bdc967d75c683396abb4b813a68fafe","last_reissued_at":"2026-07-05T03:26:42.753644Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:26:42.753644Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Combiner: Full Attention Transformer with Sparse Computation Cost","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.CV"],"primary_cat":"cs.LG","authors_text":"Bo Dai, Dale Schuurmans, Hanjun Dai, Hongyu Ren, Jure Leskovec, Mengjiao Yang, Zihang Dai","submitted_at":"2021-07-12T22:43:11Z","abstract_excerpt":"Transformers provide a class of expressive architectures that are extremely effective for sequence modeling. However, the key limitation of transformers is their quadratic memory and time complexity $\\mathcal{O}(L^2)$ with respect to the sequence length in attention layers, which restricts application in extremely long sequences. Most existing approaches leverage sparsity or low-rank assumptions in the attention matrix to reduce cost, but sacrifice expressiveness. Instead, we propose Combiner, which provides full attention capability in each attention head while maintaining low computation and"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2107.05768","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/2107.05768/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":"2107.05768","created_at":"2026-07-05T03:26:42.753704+00:00"},{"alias_kind":"arxiv_version","alias_value":"2107.05768v2","created_at":"2026-07-05T03:26:42.753704+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2107.05768","created_at":"2026-07-05T03:26:42.753704+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZPDLBDPHLR45","created_at":"2026-07-05T03:26:42.753704+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZPDLBDPHLR45TNZW","created_at":"2026-07-05T03:26:42.753704+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZPDLBDPH","created_at":"2026-07-05T03:26:42.753704+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/ZPDLBDPHLR45TNZWU5YF4JDNZV","json":"https://pith.science/pith/ZPDLBDPHLR45TNZWU5YF4JDNZV.json","graph_json":"https://pith.science/api/pith-number/ZPDLBDPHLR45TNZWU5YF4JDNZV/graph.json","events_json":"https://pith.science/api/pith-number/ZPDLBDPHLR45TNZWU5YF4JDNZV/events.json","paper":"https://pith.science/paper/ZPDLBDPH"},"agent_actions":{"view_html":"https://pith.science/pith/ZPDLBDPHLR45TNZWU5YF4JDNZV","download_json":"https://pith.science/pith/ZPDLBDPHLR45TNZWU5YF4JDNZV.json","view_paper":"https://pith.science/paper/ZPDLBDPH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2107.05768&json=true","fetch_graph":"https://pith.science/api/pith-number/ZPDLBDPHLR45TNZWU5YF4JDNZV/graph.json","fetch_events":"https://pith.science/api/pith-number/ZPDLBDPHLR45TNZWU5YF4JDNZV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZPDLBDPHLR45TNZWU5YF4JDNZV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZPDLBDPHLR45TNZWU5YF4JDNZV/action/storage_attestation","attest_author":"https://pith.science/pith/ZPDLBDPHLR45TNZWU5YF4JDNZV/action/author_attestation","sign_citation":"https://pith.science/pith/ZPDLBDPHLR45TNZWU5YF4JDNZV/action/citation_signature","submit_replication":"https://pith.science/pith/ZPDLBDPHLR45TNZWU5YF4JDNZV/action/replication_record"}},"created_at":"2026-07-05T03:26:42.753704+00:00","updated_at":"2026-07-05T03:26:42.753704+00:00"}