{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:DGQZNJEEWDOUZYBQWAG5GFCWCI","short_pith_number":"pith:DGQZNJEE","schema_version":"1.0","canonical_sha256":"19a196a484b0dd4ce030b00dd31456121f152ee524e41306fac20dafdb06c07c","source":{"kind":"arxiv","id":"2406.00048","version":3},"attestation_state":"computed","paper":{"title":"Towards a theory of how the structure of language is acquired by deep neural networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.LG"],"primary_cat":"cs.CL","authors_text":"Francesco Cagnetta, Matthieu Wyart","submitted_at":"2024-05-28T17:01:22Z","abstract_excerpt":"How much data is required to learn the structure of a language via next-token prediction? We study this question for synthetic datasets generated via a Probabilistic Context-Free Grammar (PCFG) -- a tree-like generative model that captures many of the hierarchical structures found in natural languages. We determine token-token correlations analytically in our model and show that they can be used to build a representation of the grammar's hidden variables, the longer the range the deeper the variable. In addition, a finite training set limits the resolution of correlations to an effective range"},"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":"2406.00048","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-05-28T17:01:22Z","cross_cats_sorted":["cond-mat.dis-nn","cs.LG"],"title_canon_sha256":"78b71155938f87dbbf6f36ad1d13925d6e2cff97b4ac4e3e125af48a3250eaea","abstract_canon_sha256":"ee10239a0ee8077ab5b7a007114fbb52e33801c0bbef1a3c7b465b166f580bf6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:27:54.725861Z","signature_b64":"BhC9dZBIXtOCqv2CDlC5awbohZK/vmG1PNNEcKdQNORem/3baznpVYCSVTn6p9dgeTZnRNGt44ez9BsbCIUaDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"19a196a484b0dd4ce030b00dd31456121f152ee524e41306fac20dafdb06c07c","last_reissued_at":"2026-07-05T09:27:54.725356Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:27:54.725356Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Towards a theory of how the structure of language is acquired by deep neural networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","cs.LG"],"primary_cat":"cs.CL","authors_text":"Francesco Cagnetta, Matthieu Wyart","submitted_at":"2024-05-28T17:01:22Z","abstract_excerpt":"How much data is required to learn the structure of a language via next-token prediction? We study this question for synthetic datasets generated via a Probabilistic Context-Free Grammar (PCFG) -- a tree-like generative model that captures many of the hierarchical structures found in natural languages. We determine token-token correlations analytically in our model and show that they can be used to build a representation of the grammar's hidden variables, the longer the range the deeper the variable. In addition, a finite training set limits the resolution of correlations to an effective range"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.00048","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/2406.00048/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":"2406.00048","created_at":"2026-07-05T09:27:54.725417+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.00048v3","created_at":"2026-07-05T09:27:54.725417+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.00048","created_at":"2026-07-05T09:27:54.725417+00:00"},{"alias_kind":"pith_short_12","alias_value":"DGQZNJEEWDOU","created_at":"2026-07-05T09:27:54.725417+00:00"},{"alias_kind":"pith_short_16","alias_value":"DGQZNJEEWDOUZYBQ","created_at":"2026-07-05T09:27:54.725417+00:00"},{"alias_kind":"pith_short_8","alias_value":"DGQZNJEE","created_at":"2026-07-05T09:27:54.725417+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20299","citing_title":"Statistical Properties of Training & Generalization","ref_index":215,"is_internal_anchor":false},{"citing_arxiv_id":"2606.20299","citing_title":"Statistical Properties of Training & Generalization","ref_index":215,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DGQZNJEEWDOUZYBQWAG5GFCWCI","json":"https://pith.science/pith/DGQZNJEEWDOUZYBQWAG5GFCWCI.json","graph_json":"https://pith.science/api/pith-number/DGQZNJEEWDOUZYBQWAG5GFCWCI/graph.json","events_json":"https://pith.science/api/pith-number/DGQZNJEEWDOUZYBQWAG5GFCWCI/events.json","paper":"https://pith.science/paper/DGQZNJEE"},"agent_actions":{"view_html":"https://pith.science/pith/DGQZNJEEWDOUZYBQWAG5GFCWCI","download_json":"https://pith.science/pith/DGQZNJEEWDOUZYBQWAG5GFCWCI.json","view_paper":"https://pith.science/paper/DGQZNJEE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.00048&json=true","fetch_graph":"https://pith.science/api/pith-number/DGQZNJEEWDOUZYBQWAG5GFCWCI/graph.json","fetch_events":"https://pith.science/api/pith-number/DGQZNJEEWDOUZYBQWAG5GFCWCI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DGQZNJEEWDOUZYBQWAG5GFCWCI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DGQZNJEEWDOUZYBQWAG5GFCWCI/action/storage_attestation","attest_author":"https://pith.science/pith/DGQZNJEEWDOUZYBQWAG5GFCWCI/action/author_attestation","sign_citation":"https://pith.science/pith/DGQZNJEEWDOUZYBQWAG5GFCWCI/action/citation_signature","submit_replication":"https://pith.science/pith/DGQZNJEEWDOUZYBQWAG5GFCWCI/action/replication_record"}},"created_at":"2026-07-05T09:27:54.725417+00:00","updated_at":"2026-07-05T09:27:54.725417+00:00"}