{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:JHKZH72TSDUALKOSEUNCVDQCSA","short_pith_number":"pith:JHKZH72T","schema_version":"1.0","canonical_sha256":"49d593ff5390e805a9d2251a2a8e029026526bdd0c4a41a6e35bee38a5cd0baf","source":{"kind":"arxiv","id":"2410.11474","version":3},"attestation_state":"computed","paper":{"title":"How Transformers Get Rich: Approximation and Dynamics Analysis","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Lei Wu, Mingze Wang, Ruoxi Yu, Weinan E","submitted_at":"2024-10-15T10:22:27Z","abstract_excerpt":"Transformers have demonstrated exceptional in-context learning capabilities, yet the theoretical understanding of the underlying mechanisms remains limited. A recent work (Elhage et al., 2021) identified a ``rich'' in-context mechanism known as induction head, contrasting with ``lazy'' $n$-gram models that overlook long-range dependencies. In this work, we provide both approximation and dynamics analyses of how transformers implement induction heads. In the {\\em approximation} analysis, we formalize both standard and generalized induction head mechanisms, and examine how transformers can effic"},"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":"2410.11474","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-10-15T10:22:27Z","cross_cats_sorted":["math.OC","stat.ML"],"title_canon_sha256":"bffb931c9ce77eaa8cf25cb105ad120b42fcd2587f33ca50b3b268bf69c6cf4a","abstract_canon_sha256":"516ba863057d351869b532b6bd25f60c1a0c5d171e08c40673c07334516a29f5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:06:44.680717Z","signature_b64":"TTb5m2lmU/p6CVqg+Hb33vM2e9GgTyqVA02EFDyPTJHAjo6GMg8HJ+2BDXPQlrRU/P2pK46LZLUNzTcPdRReCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"49d593ff5390e805a9d2251a2a8e029026526bdd0c4a41a6e35bee38a5cd0baf","last_reissued_at":"2026-07-05T10:06:44.680179Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:06:44.680179Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Transformers Get Rich: Approximation and Dynamics Analysis","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Lei Wu, Mingze Wang, Ruoxi Yu, Weinan E","submitted_at":"2024-10-15T10:22:27Z","abstract_excerpt":"Transformers have demonstrated exceptional in-context learning capabilities, yet the theoretical understanding of the underlying mechanisms remains limited. A recent work (Elhage et al., 2021) identified a ``rich'' in-context mechanism known as induction head, contrasting with ``lazy'' $n$-gram models that overlook long-range dependencies. In this work, we provide both approximation and dynamics analyses of how transformers implement induction heads. In the {\\em approximation} analysis, we formalize both standard and generalized induction head mechanisms, and examine how transformers can effic"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.11474","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/2410.11474/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":"2410.11474","created_at":"2026-07-05T10:06:44.680241+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.11474v3","created_at":"2026-07-05T10:06:44.680241+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.11474","created_at":"2026-07-05T10:06:44.680241+00:00"},{"alias_kind":"pith_short_12","alias_value":"JHKZH72TSDUA","created_at":"2026-07-05T10:06:44.680241+00:00"},{"alias_kind":"pith_short_16","alias_value":"JHKZH72TSDUALKOS","created_at":"2026-07-05T10:06:44.680241+00:00"},{"alias_kind":"pith_short_8","alias_value":"JHKZH72T","created_at":"2026-07-05T10:06:44.680241+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26749","citing_title":"Structure Before Collapse: Transient semantic geometry in next-token prediction","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18164","citing_title":"Learning Dynamics of Chain-of-Thought State Tracking in a Solvable Transformer Model","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12058","citing_title":"Phase Transitions in Attention: A Bayesian Theory of Copy Head Emergence","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04662","citing_title":"Why Muon Outperforms Adam: A Curvature Perspective","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2601.19208","citing_title":"How Do Transformers Learn to Associate Tokens: Gradient Leading Terms Bring Mechanistic Interpretability","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JHKZH72TSDUALKOSEUNCVDQCSA","json":"https://pith.science/pith/JHKZH72TSDUALKOSEUNCVDQCSA.json","graph_json":"https://pith.science/api/pith-number/JHKZH72TSDUALKOSEUNCVDQCSA/graph.json","events_json":"https://pith.science/api/pith-number/JHKZH72TSDUALKOSEUNCVDQCSA/events.json","paper":"https://pith.science/paper/JHKZH72T"},"agent_actions":{"view_html":"https://pith.science/pith/JHKZH72TSDUALKOSEUNCVDQCSA","download_json":"https://pith.science/pith/JHKZH72TSDUALKOSEUNCVDQCSA.json","view_paper":"https://pith.science/paper/JHKZH72T","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.11474&json=true","fetch_graph":"https://pith.science/api/pith-number/JHKZH72TSDUALKOSEUNCVDQCSA/graph.json","fetch_events":"https://pith.science/api/pith-number/JHKZH72TSDUALKOSEUNCVDQCSA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JHKZH72TSDUALKOSEUNCVDQCSA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JHKZH72TSDUALKOSEUNCVDQCSA/action/storage_attestation","attest_author":"https://pith.science/pith/JHKZH72TSDUALKOSEUNCVDQCSA/action/author_attestation","sign_citation":"https://pith.science/pith/JHKZH72TSDUALKOSEUNCVDQCSA/action/citation_signature","submit_replication":"https://pith.science/pith/JHKZH72TSDUALKOSEUNCVDQCSA/action/replication_record"}},"created_at":"2026-07-05T10:06:44.680241+00:00","updated_at":"2026-07-05T10:06:44.680241+00:00"}