{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2019:34UDLN3QAFUQLPKZXSRPPS3MHD","short_pith_number":"pith:34UDLN3Q","schema_version":"1.0","canonical_sha256":"df2835b770016905bd59bca2f7cb6c38fb812e8d845b85d92b97a3d344f3b077","source":{"kind":"arxiv","id":"1906.02762","version":1},"attestation_state":"computed","paper":{"title":"Understanding and Improving Transformer From a Multi-Particle Dynamic System Point of View","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Bin Dong, Di He, Liwei Wang, Tao Qin, Tie-Yan Liu, Yiping Lu, Zhiqing Sun, Zhuohan Li","submitted_at":"2019-06-06T18:10:08Z","abstract_excerpt":"The Transformer architecture is widely used in natural language processing. Despite its success, the design principle of the Transformer remains elusive. In this paper, we provide a novel perspective towards understanding the architecture: we show that the Transformer can be mathematically interpreted as a numerical Ordinary Differential Equation (ODE) solver for a convection-diffusion equation in a multi-particle dynamic system. In particular, how words in a sentence are abstracted into contexts by passing through the layers of the Transformer can be interpreted as approximating multiple part"},"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":"1906.02762","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2019-06-06T18:10:08Z","cross_cats_sorted":["cs.CL","stat.ML"],"title_canon_sha256":"178b6de7b9593d3e0b8092fe59072f78591226ef15b6752ccbb5e04e74755364","abstract_canon_sha256":"8ade44b469ae0ae3d33aadbe114b80a6f132616e1605e1fbd17744bf4e64ec36"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:43:59.856924Z","signature_b64":"zPm2CkgQkQlSy8f5VABsHR+4dyeh4GWWPTImQfQjmKokQxkweuDpYtBTFbAn/kop0tVz4a8y9QCabiG22QW6AA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"df2835b770016905bd59bca2f7cb6c38fb812e8d845b85d92b97a3d344f3b077","last_reissued_at":"2026-05-17T23:43:59.856345Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:43:59.856345Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Understanding and Improving Transformer From a Multi-Particle Dynamic System Point of View","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","stat.ML"],"primary_cat":"cs.LG","authors_text":"Bin Dong, Di He, Liwei Wang, Tao Qin, Tie-Yan Liu, Yiping Lu, Zhiqing Sun, Zhuohan Li","submitted_at":"2019-06-06T18:10:08Z","abstract_excerpt":"The Transformer architecture is widely used in natural language processing. Despite its success, the design principle of the Transformer remains elusive. In this paper, we provide a novel perspective towards understanding the architecture: we show that the Transformer can be mathematically interpreted as a numerical Ordinary Differential Equation (ODE) solver for a convection-diffusion equation in a multi-particle dynamic system. In particular, how words in a sentence are abstracted into contexts by passing through the layers of the Transformer can be interpreted as approximating multiple part"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1906.02762","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"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":"1906.02762","created_at":"2026-05-17T23:43:59.856429+00:00"},{"alias_kind":"arxiv_version","alias_value":"1906.02762v1","created_at":"2026-05-17T23:43:59.856429+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1906.02762","created_at":"2026-05-17T23:43:59.856429+00:00"},{"alias_kind":"pith_short_12","alias_value":"34UDLN3QAFUQ","created_at":"2026-05-18T12:33:07.085635+00:00"},{"alias_kind":"pith_short_16","alias_value":"34UDLN3QAFUQLPKZ","created_at":"2026-05-18T12:33:07.085635+00:00"},{"alias_kind":"pith_short_8","alias_value":"34UDLN3Q","created_at":"2026-05-18T12:33:07.085635+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.14258","citing_title":"Dynamics of the Transformer Residual Stream: Coupling Spectral Geometry to Network Topology","ref_index":15,"is_internal_anchor":true},{"citing_arxiv_id":"2605.01374","citing_title":"MTA: Multi-Granular Trajectory Alignment for Large Language Model Distillation","ref_index":47,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01205","citing_title":"SRA: Span Representation Alignment for Large Language Model Distillation","ref_index":47,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/34UDLN3QAFUQLPKZXSRPPS3MHD","json":"https://pith.science/pith/34UDLN3QAFUQLPKZXSRPPS3MHD.json","graph_json":"https://pith.science/api/pith-number/34UDLN3QAFUQLPKZXSRPPS3MHD/graph.json","events_json":"https://pith.science/api/pith-number/34UDLN3QAFUQLPKZXSRPPS3MHD/events.json","paper":"https://pith.science/paper/34UDLN3Q"},"agent_actions":{"view_html":"https://pith.science/pith/34UDLN3QAFUQLPKZXSRPPS3MHD","download_json":"https://pith.science/pith/34UDLN3QAFUQLPKZXSRPPS3MHD.json","view_paper":"https://pith.science/paper/34UDLN3Q","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1906.02762&json=true","fetch_graph":"https://pith.science/api/pith-number/34UDLN3QAFUQLPKZXSRPPS3MHD/graph.json","fetch_events":"https://pith.science/api/pith-number/34UDLN3QAFUQLPKZXSRPPS3MHD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/34UDLN3QAFUQLPKZXSRPPS3MHD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/34UDLN3QAFUQLPKZXSRPPS3MHD/action/storage_attestation","attest_author":"https://pith.science/pith/34UDLN3QAFUQLPKZXSRPPS3MHD/action/author_attestation","sign_citation":"https://pith.science/pith/34UDLN3QAFUQLPKZXSRPPS3MHD/action/citation_signature","submit_replication":"https://pith.science/pith/34UDLN3QAFUQLPKZXSRPPS3MHD/action/replication_record"}},"created_at":"2026-05-17T23:43:59.856429+00:00","updated_at":"2026-05-17T23:43:59.856429+00:00"}