{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:EF4RPSA7JWEBX7SBVY4MGDA3VR","short_pith_number":"pith:EF4RPSA7","schema_version":"1.0","canonical_sha256":"217917c81f4d881bfe41ae38c30c1bac4409db4f12ee88b0bd102f50620d204e","source":{"kind":"arxiv","id":"2405.15926","version":2},"attestation_state":"computed","paper":{"title":"Dissecting the Interplay of Attention Paths in a Statistical Mechanics Theory of Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cond-mat.dis-nn","cond-mat.stat-mech","stat.ML"],"primary_cat":"cs.LG","authors_text":"Francesca Mignacco, Haim Sompolinsky, Kazuki Irie, Lorenzo Tiberi","submitted_at":"2024-05-24T20:34:18Z","abstract_excerpt":"Despite the remarkable empirical performance of Transformers, their theoretical understanding remains elusive. Here, we consider a deep multi-head self-attention network, that is closely related to Transformers yet analytically tractable. We develop a statistical mechanics theory of Bayesian learning in this model, deriving exact equations for the network's predictor statistics under the finite-width thermodynamic limit, i.e., $N,P\\rightarrow\\infty$, $P/N=\\mathcal{O}(1)$, where $N$ is the network width and $P$ is the number of training examples. Our theory shows that the predictor statistics 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":"2405.15926","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-05-24T20:34:18Z","cross_cats_sorted":["cond-mat.dis-nn","cond-mat.stat-mech","stat.ML"],"title_canon_sha256":"e1fd643378b4c362886b0e72f513381309e7ef68c780bbb32cc7ccbb0a3b9f1f","abstract_canon_sha256":"98ddbe5e17fe35ad6f5f2e90ea9840e93b5326f30c7a3287ec68e7172c891980"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:45:49.306622Z","signature_b64":"u1oJNWHKd0r2JjDiweLvOZ4Wnu1dX+9we/P1KkPGHS5Qpsih8oZDZ5rAK88t+zC8q+ECASlgIoxnWIWfef9eBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"217917c81f4d881bfe41ae38c30c1bac4409db4f12ee88b0bd102f50620d204e","last_reissued_at":"2026-07-05T09:45:49.306111Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:45:49.306111Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Dissecting the Interplay of Attention Paths in a Statistical Mechanics Theory of Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cond-mat.dis-nn","cond-mat.stat-mech","stat.ML"],"primary_cat":"cs.LG","authors_text":"Francesca Mignacco, Haim Sompolinsky, Kazuki Irie, Lorenzo Tiberi","submitted_at":"2024-05-24T20:34:18Z","abstract_excerpt":"Despite the remarkable empirical performance of Transformers, their theoretical understanding remains elusive. Here, we consider a deep multi-head self-attention network, that is closely related to Transformers yet analytically tractable. We develop a statistical mechanics theory of Bayesian learning in this model, deriving exact equations for the network's predictor statistics under the finite-width thermodynamic limit, i.e., $N,P\\rightarrow\\infty$, $P/N=\\mathcal{O}(1)$, where $N$ is the network width and $P$ is the number of training examples. Our theory shows that the predictor statistics a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.15926","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/2405.15926/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":"2405.15926","created_at":"2026-07-05T09:45:49.306172+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.15926v2","created_at":"2026-07-05T09:45:49.306172+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.15926","created_at":"2026-07-05T09:45:49.306172+00:00"},{"alias_kind":"pith_short_12","alias_value":"EF4RPSA7JWEB","created_at":"2026-07-05T09:45:49.306172+00:00"},{"alias_kind":"pith_short_16","alias_value":"EF4RPSA7JWEBX7SB","created_at":"2026-07-05T09:45:49.306172+00:00"},{"alias_kind":"pith_short_8","alias_value":"EF4RPSA7","created_at":"2026-07-05T09:45:49.306172+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/EF4RPSA7JWEBX7SBVY4MGDA3VR","json":"https://pith.science/pith/EF4RPSA7JWEBX7SBVY4MGDA3VR.json","graph_json":"https://pith.science/api/pith-number/EF4RPSA7JWEBX7SBVY4MGDA3VR/graph.json","events_json":"https://pith.science/api/pith-number/EF4RPSA7JWEBX7SBVY4MGDA3VR/events.json","paper":"https://pith.science/paper/EF4RPSA7"},"agent_actions":{"view_html":"https://pith.science/pith/EF4RPSA7JWEBX7SBVY4MGDA3VR","download_json":"https://pith.science/pith/EF4RPSA7JWEBX7SBVY4MGDA3VR.json","view_paper":"https://pith.science/paper/EF4RPSA7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.15926&json=true","fetch_graph":"https://pith.science/api/pith-number/EF4RPSA7JWEBX7SBVY4MGDA3VR/graph.json","fetch_events":"https://pith.science/api/pith-number/EF4RPSA7JWEBX7SBVY4MGDA3VR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EF4RPSA7JWEBX7SBVY4MGDA3VR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EF4RPSA7JWEBX7SBVY4MGDA3VR/action/storage_attestation","attest_author":"https://pith.science/pith/EF4RPSA7JWEBX7SBVY4MGDA3VR/action/author_attestation","sign_citation":"https://pith.science/pith/EF4RPSA7JWEBX7SBVY4MGDA3VR/action/citation_signature","submit_replication":"https://pith.science/pith/EF4RPSA7JWEBX7SBVY4MGDA3VR/action/replication_record"}},"created_at":"2026-07-05T09:45:49.306172+00:00","updated_at":"2026-07-05T09:45:49.306172+00:00"}