{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:22FWHIAWSPJIT25RQ3KVQ4OTY4","short_pith_number":"pith:22FWHIAW","schema_version":"1.0","canonical_sha256":"d68b63a01693d289ebb186d55871d3c732f60b47281326ffabce3226a217f465","source":{"kind":"arxiv","id":"2402.04362","version":3},"attestation_state":"computed","paper":{"title":"Neural Networks Learn Statistics of Increasing Complexity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alex Mallen, Lucia Quirke, Nora Belrose, Quintin Pope, Xiaoli Fern","submitted_at":"2024-02-06T20:03:35Z","abstract_excerpt":"The distributional simplicity bias (DSB) posits that neural networks learn low-order moments of the data distribution first, before moving on to higher-order correlations. In this work, we present compelling new evidence for the DSB by showing that networks automatically learn to perform well on maximum-entropy distributions whose low-order statistics match those of the training set early in training, then lose this ability later. We also extend the DSB to discrete domains by proving an equivalence between token $n$-gram frequencies and the moments of embedding vectors, and by finding empirica"},"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":"2402.04362","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-06T20:03:35Z","cross_cats_sorted":[],"title_canon_sha256":"116fcc7a1f64057d17d307555b36987995c6d0a00af141718865f4b92355c815","abstract_canon_sha256":"7c13b951147d91e66d435bf03c8f6b0841a7d91229496cb319522886cdc437da"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:17:37.134898Z","signature_b64":"0WvrIoKdvHpjjZK908ainB3mFdl+kFIQdkwJGUPrXVrifxS7UWxijJGCZ2BN19bl1b1FRU1ytRVzRLC4eQ8EDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d68b63a01693d289ebb186d55871d3c732f60b47281326ffabce3226a217f465","last_reissued_at":"2026-07-05T09:17:37.134372Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:17:37.134372Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Neural Networks Learn Statistics of Increasing Complexity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alex Mallen, Lucia Quirke, Nora Belrose, Quintin Pope, Xiaoli Fern","submitted_at":"2024-02-06T20:03:35Z","abstract_excerpt":"The distributional simplicity bias (DSB) posits that neural networks learn low-order moments of the data distribution first, before moving on to higher-order correlations. In this work, we present compelling new evidence for the DSB by showing that networks automatically learn to perform well on maximum-entropy distributions whose low-order statistics match those of the training set early in training, then lose this ability later. We also extend the DSB to discrete domains by proving an equivalence between token $n$-gram frequencies and the moments of embedding vectors, and by finding empirica"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.04362","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/2402.04362/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":"2402.04362","created_at":"2026-07-05T09:17:37.134442+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.04362v3","created_at":"2026-07-05T09:17:37.134442+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.04362","created_at":"2026-07-05T09:17:37.134442+00:00"},{"alias_kind":"pith_short_12","alias_value":"22FWHIAWSPJI","created_at":"2026-07-05T09:17:37.134442+00:00"},{"alias_kind":"pith_short_16","alias_value":"22FWHIAWSPJIT25R","created_at":"2026-07-05T09:17:37.134442+00:00"},{"alias_kind":"pith_short_8","alias_value":"22FWHIAW","created_at":"2026-07-05T09:17:37.134442+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2603.12901","citing_title":"A theory of learning data statistics in diffusion models, from easy to hard","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15183","citing_title":"When Are Two Networks the Same? Tensor Similarity for Mechanistic Interpretability","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07844","citing_title":"Distributional simplicity bias and effective convexity in Energy Based Models","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/22FWHIAWSPJIT25RQ3KVQ4OTY4","json":"https://pith.science/pith/22FWHIAWSPJIT25RQ3KVQ4OTY4.json","graph_json":"https://pith.science/api/pith-number/22FWHIAWSPJIT25RQ3KVQ4OTY4/graph.json","events_json":"https://pith.science/api/pith-number/22FWHIAWSPJIT25RQ3KVQ4OTY4/events.json","paper":"https://pith.science/paper/22FWHIAW"},"agent_actions":{"view_html":"https://pith.science/pith/22FWHIAWSPJIT25RQ3KVQ4OTY4","download_json":"https://pith.science/pith/22FWHIAWSPJIT25RQ3KVQ4OTY4.json","view_paper":"https://pith.science/paper/22FWHIAW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.04362&json=true","fetch_graph":"https://pith.science/api/pith-number/22FWHIAWSPJIT25RQ3KVQ4OTY4/graph.json","fetch_events":"https://pith.science/api/pith-number/22FWHIAWSPJIT25RQ3KVQ4OTY4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/22FWHIAWSPJIT25RQ3KVQ4OTY4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/22FWHIAWSPJIT25RQ3KVQ4OTY4/action/storage_attestation","attest_author":"https://pith.science/pith/22FWHIAWSPJIT25RQ3KVQ4OTY4/action/author_attestation","sign_citation":"https://pith.science/pith/22FWHIAWSPJIT25RQ3KVQ4OTY4/action/citation_signature","submit_replication":"https://pith.science/pith/22FWHIAWSPJIT25RQ3KVQ4OTY4/action/replication_record"}},"created_at":"2026-07-05T09:17:37.134442+00:00","updated_at":"2026-07-05T09:17:37.134442+00:00"}