{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:D2N46AEYTAUNUVIJWHC7PVABLY","short_pith_number":"pith:D2N46AEY","schema_version":"1.0","canonical_sha256":"1e9bcf00989828da5509b1c5f7d4015e1245702bf928be3a61a7b1fa95623afb","source":{"kind":"arxiv","id":"2206.10012","version":1},"attestation_state":"computed","paper":{"title":"Limitations of the NTK for Understanding Generalization in Deep Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Nikhil Vyas, Preetum Nakkiran, Yamini Bansal","submitted_at":"2022-06-20T21:23:28Z","abstract_excerpt":"The ``Neural Tangent Kernel'' (NTK) (Jacot et al 2018), and its empirical variants have been proposed as a proxy to capture certain behaviors of real neural networks. In this work, we study NTKs through the lens of scaling laws, and demonstrate that they fall short of explaining important aspects of neural network generalization. In particular, we demonstrate realistic settings where finite-width neural networks have significantly better data scaling exponents as compared to their corresponding empirical and infinite NTKs at initialization. This reveals a more fundamental difference between th"},"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":"2206.10012","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2022-06-20T21:23:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"53cc5302fcd6b258ef6ac1a5a78679beff00efcfc4bc5172d92c8122a6b19eba","abstract_canon_sha256":"22edde95a04f7233f6b6d35aa8e809c7a99cfad35798d8d4b68a073be581e2c7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:33:42.373695Z","signature_b64":"t74/+BiyXO4PPl2SOw3HCQkErH/jLDHZcN93TLSuB5N9Q7MqnagNOuNdPcTxLcWGLYR0/s+rPjlWywsL0X4jAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1e9bcf00989828da5509b1c5f7d4015e1245702bf928be3a61a7b1fa95623afb","last_reissued_at":"2026-07-05T04:33:42.373216Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:33:42.373216Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Limitations of the NTK for Understanding Generalization in Deep Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Nikhil Vyas, Preetum Nakkiran, Yamini Bansal","submitted_at":"2022-06-20T21:23:28Z","abstract_excerpt":"The ``Neural Tangent Kernel'' (NTK) (Jacot et al 2018), and its empirical variants have been proposed as a proxy to capture certain behaviors of real neural networks. In this work, we study NTKs through the lens of scaling laws, and demonstrate that they fall short of explaining important aspects of neural network generalization. In particular, we demonstrate realistic settings where finite-width neural networks have significantly better data scaling exponents as compared to their corresponding empirical and infinite NTKs at initialization. This reveals a more fundamental difference between th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2206.10012","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/2206.10012/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":"2206.10012","created_at":"2026-07-05T04:33:42.373272+00:00"},{"alias_kind":"arxiv_version","alias_value":"2206.10012v1","created_at":"2026-07-05T04:33:42.373272+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2206.10012","created_at":"2026-07-05T04:33:42.373272+00:00"},{"alias_kind":"pith_short_12","alias_value":"D2N46AEYTAUN","created_at":"2026-07-05T04:33:42.373272+00:00"},{"alias_kind":"pith_short_16","alias_value":"D2N46AEYTAUNUVIJ","created_at":"2026-07-05T04:33:42.373272+00:00"},{"alias_kind":"pith_short_8","alias_value":"D2N46AEY","created_at":"2026-07-05T04:33:42.373272+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07495","citing_title":"Second-Order Path Kernel Interpolation Formulas in Machine Learning","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17767","citing_title":"Feature Learning in Linear-Width Two-Layer Networks: Two vs. One Step of Gradient Descent","ref_index":232,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07870","citing_title":"Spectral Dynamics in Deep Networks: Feature Learning, Outlier Escape, and Learning Rate Transfer","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18598","citing_title":"Pointwise Generalization in Deep Neural Networks","ref_index":95,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17767","citing_title":"Feature Learning in Linear-Width Two-Layer Networks: Two vs. One Step of Gradient Descent","ref_index":232,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14200","citing_title":"How to Scale Mixture-of-Experts: From muP to the Maximally Scale-Stable Parameterization","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21691","citing_title":"There Will Be a Scientific Theory of Deep Learning","ref_index":183,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07870","citing_title":"Spectral Dynamics in Deep Networks: Feature Learning, Outlier Escape, and Learning Rate Transfer","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/D2N46AEYTAUNUVIJWHC7PVABLY","json":"https://pith.science/pith/D2N46AEYTAUNUVIJWHC7PVABLY.json","graph_json":"https://pith.science/api/pith-number/D2N46AEYTAUNUVIJWHC7PVABLY/graph.json","events_json":"https://pith.science/api/pith-number/D2N46AEYTAUNUVIJWHC7PVABLY/events.json","paper":"https://pith.science/paper/D2N46AEY"},"agent_actions":{"view_html":"https://pith.science/pith/D2N46AEYTAUNUVIJWHC7PVABLY","download_json":"https://pith.science/pith/D2N46AEYTAUNUVIJWHC7PVABLY.json","view_paper":"https://pith.science/paper/D2N46AEY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2206.10012&json=true","fetch_graph":"https://pith.science/api/pith-number/D2N46AEYTAUNUVIJWHC7PVABLY/graph.json","fetch_events":"https://pith.science/api/pith-number/D2N46AEYTAUNUVIJWHC7PVABLY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D2N46AEYTAUNUVIJWHC7PVABLY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D2N46AEYTAUNUVIJWHC7PVABLY/action/storage_attestation","attest_author":"https://pith.science/pith/D2N46AEYTAUNUVIJWHC7PVABLY/action/author_attestation","sign_citation":"https://pith.science/pith/D2N46AEYTAUNUVIJWHC7PVABLY/action/citation_signature","submit_replication":"https://pith.science/pith/D2N46AEYTAUNUVIJWHC7PVABLY/action/replication_record"}},"created_at":"2026-07-05T04:33:42.373272+00:00","updated_at":"2026-07-05T04:33:42.373272+00:00"}