{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:Z3GDLMSLBNJPTYOEHJTMEM6OR3","short_pith_number":"pith:Z3GDLMSL","schema_version":"1.0","canonical_sha256":"cecc35b24b0b52f9e1c43a66c233ce8ece06d54db6da6221a5ebab0582f1a997","source":{"kind":"arxiv","id":"2505.07070","version":1},"attestation_state":"computed","paper":{"title":"Scaling Laws and Representation Learning in Simple Hierarchical Languages: Transformers vs. Convolutional Architectures","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","stat.ML"],"primary_cat":"cs.LG","authors_text":"Alessandro Favero, Antonio Sclocchi, Francesco Cagnetta, Matthieu Wyart","submitted_at":"2025-05-11T17:44:14Z","abstract_excerpt":"How do neural language models acquire a language's structure when trained for next-token prediction? We address this question by deriving theoretical scaling laws for neural network performance on synthetic datasets generated by the Random Hierarchy Model (RHM) -- an ensemble of probabilistic context-free grammars designed to capture the hierarchical structure of natural language while remaining analytically tractable. Previously, we developed a theory of representation learning based on data correlations that explains how deep learning models capture the hierarchical structure of the data seq"},"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":"2505.07070","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-11T17:44:14Z","cross_cats_sorted":["cond-mat.dis-nn","stat.ML"],"title_canon_sha256":"b4fcf3c360d0d38809da7c657ba9cdb503b321a78a947d9212728e6e0cb5be5e","abstract_canon_sha256":"d17e768028433e822880754f1df281143f0e4aa6bd304468fd16f308297f394f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:01:42.254614Z","signature_b64":"N0fUCBu12PDY7o1SQHNfLeqkKZNzud4EE0R0hHoG0HJ997ghPKEZe+eKeH0fFORo5PXvPSTVMcEI4CkN2ThlBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cecc35b24b0b52f9e1c43a66c233ce8ece06d54db6da6221a5ebab0582f1a997","last_reissued_at":"2026-07-05T11:01:42.254152Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:01:42.254152Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Scaling Laws and Representation Learning in Simple Hierarchical Languages: Transformers vs. Convolutional Architectures","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cond-mat.dis-nn","stat.ML"],"primary_cat":"cs.LG","authors_text":"Alessandro Favero, Antonio Sclocchi, Francesco Cagnetta, Matthieu Wyart","submitted_at":"2025-05-11T17:44:14Z","abstract_excerpt":"How do neural language models acquire a language's structure when trained for next-token prediction? We address this question by deriving theoretical scaling laws for neural network performance on synthetic datasets generated by the Random Hierarchy Model (RHM) -- an ensemble of probabilistic context-free grammars designed to capture the hierarchical structure of natural language while remaining analytically tractable. Previously, we developed a theory of representation learning based on data correlations that explains how deep learning models capture the hierarchical structure of the data seq"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.07070","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":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2505.07070/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":"2505.07070","created_at":"2026-07-05T11:01:42.254210+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.07070v1","created_at":"2026-07-05T11:01:42.254210+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.07070","created_at":"2026-07-05T11:01:42.254210+00:00"},{"alias_kind":"pith_short_12","alias_value":"Z3GDLMSLBNJP","created_at":"2026-07-05T11:01:42.254210+00:00"},{"alias_kind":"pith_short_16","alias_value":"Z3GDLMSLBNJPTYOE","created_at":"2026-07-05T11:01:42.254210+00:00"},{"alias_kind":"pith_short_8","alias_value":"Z3GDLMSL","created_at":"2026-07-05T11:01:42.254210+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.29548","citing_title":"Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention","ref_index":40,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Z3GDLMSLBNJPTYOEHJTMEM6OR3","json":"https://pith.science/pith/Z3GDLMSLBNJPTYOEHJTMEM6OR3.json","graph_json":"https://pith.science/api/pith-number/Z3GDLMSLBNJPTYOEHJTMEM6OR3/graph.json","events_json":"https://pith.science/api/pith-number/Z3GDLMSLBNJPTYOEHJTMEM6OR3/events.json","paper":"https://pith.science/paper/Z3GDLMSL"},"agent_actions":{"view_html":"https://pith.science/pith/Z3GDLMSLBNJPTYOEHJTMEM6OR3","download_json":"https://pith.science/pith/Z3GDLMSLBNJPTYOEHJTMEM6OR3.json","view_paper":"https://pith.science/paper/Z3GDLMSL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.07070&json=true","fetch_graph":"https://pith.science/api/pith-number/Z3GDLMSLBNJPTYOEHJTMEM6OR3/graph.json","fetch_events":"https://pith.science/api/pith-number/Z3GDLMSLBNJPTYOEHJTMEM6OR3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Z3GDLMSLBNJPTYOEHJTMEM6OR3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Z3GDLMSLBNJPTYOEHJTMEM6OR3/action/storage_attestation","attest_author":"https://pith.science/pith/Z3GDLMSLBNJPTYOEHJTMEM6OR3/action/author_attestation","sign_citation":"https://pith.science/pith/Z3GDLMSLBNJPTYOEHJTMEM6OR3/action/citation_signature","submit_replication":"https://pith.science/pith/Z3GDLMSLBNJPTYOEHJTMEM6OR3/action/replication_record"}},"created_at":"2026-07-05T11:01:42.254210+00:00","updated_at":"2026-07-05T11:01:42.254210+00:00"}