{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:RFKC2RG4HRT6UFXZ2WHPSDBAOU","short_pith_number":"pith:RFKC2RG4","schema_version":"1.0","canonical_sha256":"89542d44dc3c67ea16f9d58ef90c20753c9042f2191a262a41831240be22442f","source":{"kind":"arxiv","id":"2412.04910","version":3},"attestation_state":"computed","paper":{"title":"Learning High-Degree Parities: The Crucial Role of the Initialization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Donald Kougang-Yombi, Elisabetta Cornacchia, Emmanuel Abbe, Jan H\\k{a}z{\\l}a","submitted_at":"2024-12-06T10:05:10Z","abstract_excerpt":"Parities have become a standard benchmark for evaluating learning algorithms. Recent works show that regular neural networks trained by gradient descent can efficiently learn degree $k$ parities on uniform inputs for constant $k$, but fail to do so when $k$ and $d-k$ grow with $d$ (here $d$ is the ambient dimension). However, the case where $k=d-O_d(1)$ (almost-full parities), including the degree $d$ parity (the full parity), has remained unsettled. This paper shows that for gradient descent on regular neural networks, learnability depends on the initial weight distribution. On one hand, the "},"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":"2412.04910","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-06T10:05:10Z","cross_cats_sorted":[],"title_canon_sha256":"32eacb4a17b3c5ea93b42cb976a8f35e622ccbacd0ba5e5b316a306c03c00fe3","abstract_canon_sha256":"a41cdc2f2ef7c3e1739bbcac17c1f8902a204bd87575141f8a0532bf87962082"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:24:28.544280Z","signature_b64":"JbgOHmng7NaELK4FWdj/BL2lk5bYpuPpgb21dUXCKF2e65aRz0T44D7PZ+PjF/eSI9B3eci7YlafS4RuDqw+DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"89542d44dc3c67ea16f9d58ef90c20753c9042f2191a262a41831240be22442f","last_reissued_at":"2026-07-05T10:24:28.543486Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:24:28.543486Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning High-Degree Parities: The Crucial Role of the Initialization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Donald Kougang-Yombi, Elisabetta Cornacchia, Emmanuel Abbe, Jan H\\k{a}z{\\l}a","submitted_at":"2024-12-06T10:05:10Z","abstract_excerpt":"Parities have become a standard benchmark for evaluating learning algorithms. Recent works show that regular neural networks trained by gradient descent can efficiently learn degree $k$ parities on uniform inputs for constant $k$, but fail to do so when $k$ and $d-k$ grow with $d$ (here $d$ is the ambient dimension). However, the case where $k=d-O_d(1)$ (almost-full parities), including the degree $d$ parity (the full parity), has remained unsettled. This paper shows that for gradient descent on regular neural networks, learnability depends on the initial weight distribution. On one hand, the "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.04910","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/2412.04910/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":"2412.04910","created_at":"2026-07-05T10:24:28.543614+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.04910v3","created_at":"2026-07-05T10:24:28.543614+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.04910","created_at":"2026-07-05T10:24:28.543614+00:00"},{"alias_kind":"pith_short_12","alias_value":"RFKC2RG4HRT6","created_at":"2026-07-05T10:24:28.543614+00:00"},{"alias_kind":"pith_short_16","alias_value":"RFKC2RG4HRT6UFXZ","created_at":"2026-07-05T10:24:28.543614+00:00"},{"alias_kind":"pith_short_8","alias_value":"RFKC2RG4","created_at":"2026-07-05T10:24:28.543614+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.02393","citing_title":"Lower Bounds for Chain-of-Thought Reasoning in Hard-Attention Transformers","ref_index":2024,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RFKC2RG4HRT6UFXZ2WHPSDBAOU","json":"https://pith.science/pith/RFKC2RG4HRT6UFXZ2WHPSDBAOU.json","graph_json":"https://pith.science/api/pith-number/RFKC2RG4HRT6UFXZ2WHPSDBAOU/graph.json","events_json":"https://pith.science/api/pith-number/RFKC2RG4HRT6UFXZ2WHPSDBAOU/events.json","paper":"https://pith.science/paper/RFKC2RG4"},"agent_actions":{"view_html":"https://pith.science/pith/RFKC2RG4HRT6UFXZ2WHPSDBAOU","download_json":"https://pith.science/pith/RFKC2RG4HRT6UFXZ2WHPSDBAOU.json","view_paper":"https://pith.science/paper/RFKC2RG4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.04910&json=true","fetch_graph":"https://pith.science/api/pith-number/RFKC2RG4HRT6UFXZ2WHPSDBAOU/graph.json","fetch_events":"https://pith.science/api/pith-number/RFKC2RG4HRT6UFXZ2WHPSDBAOU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RFKC2RG4HRT6UFXZ2WHPSDBAOU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RFKC2RG4HRT6UFXZ2WHPSDBAOU/action/storage_attestation","attest_author":"https://pith.science/pith/RFKC2RG4HRT6UFXZ2WHPSDBAOU/action/author_attestation","sign_citation":"https://pith.science/pith/RFKC2RG4HRT6UFXZ2WHPSDBAOU/action/citation_signature","submit_replication":"https://pith.science/pith/RFKC2RG4HRT6UFXZ2WHPSDBAOU/action/replication_record"}},"created_at":"2026-07-05T10:24:28.543614+00:00","updated_at":"2026-07-05T10:24:28.543614+00:00"}