{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:BDF23V2IKGEBEXDPML2GRNO7TA","short_pith_number":"pith:BDF23V2I","schema_version":"1.0","canonical_sha256":"08cbadd7485188125c6f62f468b5df983ce83e8d8d1437391562fc251817f58d","source":{"kind":"arxiv","id":"2005.00743","version":3},"attestation_state":"computed","paper":{"title":"Synthesizer: Rethinking Self-Attention in Transformer Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.IR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Che Zheng, Da-Cheng Juan, Dara Bahri, Donald Metzler, Yi Tay, Zhe Zhao","submitted_at":"2020-05-02T08:16:19Z","abstract_excerpt":"The dot product self-attention is known to be central and indispensable to state-of-the-art Transformer models. But is it really required? This paper investigates the true importance and contribution of the dot product-based self-attention mechanism on the performance of Transformer models. Via extensive experiments, we find that (1) random alignment matrices surprisingly perform quite competitively and (2) learning attention weights from token-token (query-key) interactions is useful but not that important after all. To this end, we propose \\textsc{Synthesizer}, a model that learns synthetic "},"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":"2005.00743","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-05-02T08:16:19Z","cross_cats_sorted":["cs.IR","cs.LG"],"title_canon_sha256":"f2da81d49fb63975091913af33fd072210ba8937101c78ccff49a94077b52f2d","abstract_canon_sha256":"182d7139f6f0407c4a133f5f5a1f26eef37aebfecd7667db664ef4ae22eae891"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:42:25.517720Z","signature_b64":"5zefNyH83wIh4TSUN4u8Cxvn+SL/BbVToERBQk9MKkLqNsrqPm/fUpkqYMgGcFwo7yBJq6ANqjLklLosqxz8Aw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"08cbadd7485188125c6f62f468b5df983ce83e8d8d1437391562fc251817f58d","last_reissued_at":"2026-07-05T02:42:25.517278Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:42:25.517278Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Synthesizer: Rethinking Self-Attention in Transformer Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.IR","cs.LG"],"primary_cat":"cs.CL","authors_text":"Che Zheng, Da-Cheng Juan, Dara Bahri, Donald Metzler, Yi Tay, Zhe Zhao","submitted_at":"2020-05-02T08:16:19Z","abstract_excerpt":"The dot product self-attention is known to be central and indispensable to state-of-the-art Transformer models. But is it really required? This paper investigates the true importance and contribution of the dot product-based self-attention mechanism on the performance of Transformer models. Via extensive experiments, we find that (1) random alignment matrices surprisingly perform quite competitively and (2) learning attention weights from token-token (query-key) interactions is useful but not that important after all. To this end, we propose \\textsc{Synthesizer}, a model that learns synthetic "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2005.00743","kind":"arxiv","version":3},"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/2005.00743/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":"2005.00743","created_at":"2026-07-05T02:42:25.517340+00:00"},{"alias_kind":"arxiv_version","alias_value":"2005.00743v3","created_at":"2026-07-05T02:42:25.517340+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2005.00743","created_at":"2026-07-05T02:42:25.517340+00:00"},{"alias_kind":"pith_short_12","alias_value":"BDF23V2IKGEB","created_at":"2026-07-05T02:42:25.517340+00:00"},{"alias_kind":"pith_short_16","alias_value":"BDF23V2IKGEBEXDP","created_at":"2026-07-05T02:42:25.517340+00:00"},{"alias_kind":"pith_short_8","alias_value":"BDF23V2I","created_at":"2026-07-05T02:42:25.517340+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.10288","citing_title":"Power law graph attention: exact generalization of scaled dot-product attention, empirical collapse at inference","ref_index":39,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BDF23V2IKGEBEXDPML2GRNO7TA","json":"https://pith.science/pith/BDF23V2IKGEBEXDPML2GRNO7TA.json","graph_json":"https://pith.science/api/pith-number/BDF23V2IKGEBEXDPML2GRNO7TA/graph.json","events_json":"https://pith.science/api/pith-number/BDF23V2IKGEBEXDPML2GRNO7TA/events.json","paper":"https://pith.science/paper/BDF23V2I"},"agent_actions":{"view_html":"https://pith.science/pith/BDF23V2IKGEBEXDPML2GRNO7TA","download_json":"https://pith.science/pith/BDF23V2IKGEBEXDPML2GRNO7TA.json","view_paper":"https://pith.science/paper/BDF23V2I","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2005.00743&json=true","fetch_graph":"https://pith.science/api/pith-number/BDF23V2IKGEBEXDPML2GRNO7TA/graph.json","fetch_events":"https://pith.science/api/pith-number/BDF23V2IKGEBEXDPML2GRNO7TA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BDF23V2IKGEBEXDPML2GRNO7TA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BDF23V2IKGEBEXDPML2GRNO7TA/action/storage_attestation","attest_author":"https://pith.science/pith/BDF23V2IKGEBEXDPML2GRNO7TA/action/author_attestation","sign_citation":"https://pith.science/pith/BDF23V2IKGEBEXDPML2GRNO7TA/action/citation_signature","submit_replication":"https://pith.science/pith/BDF23V2IKGEBEXDPML2GRNO7TA/action/replication_record"}},"created_at":"2026-07-05T02:42:25.517340+00:00","updated_at":"2026-07-05T02:42:25.517340+00:00"}