{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FB5CS7X2UQ25HNJ2GWAGXZ2INV","short_pith_number":"pith:FB5CS7X2","schema_version":"1.0","canonical_sha256":"287a297efaa435d3b53a35806be7486d5ce9241c2dd0b61d2c317a49ede192d9","source":{"kind":"arxiv","id":"2405.20233","version":2},"attestation_state":"computed","paper":{"title":"Grokfast: Accelerated Grokking by Amplifying Slow Gradients","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bong Gyun Kang, Jaerin Lee, Kihoon Kim, Kyoung Mu Lee","submitted_at":"2024-05-30T16:35:30Z","abstract_excerpt":"One puzzling artifact in machine learning dubbed grokking is where delayed generalization is achieved tenfolds of iterations after near perfect overfitting to the training data. Focusing on the long delay itself on behalf of machine learning practitioners, our goal is to accelerate generalization of a model under grokking phenomenon. By regarding a series of gradients of a parameter over training iterations as a random signal over time, we can spectrally decompose the parameter trajectories under gradient descent into two components: the fast-varying, overfitting-yielding component and the slo"},"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":"2405.20233","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-05-30T16:35:30Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e1b5b7710aa48b7649525f317808e5ea869b26df6f8686d1e3a9e6635d617707","abstract_canon_sha256":"9ca7332774b129a3c4147682376baf6ee086551d4c50a097b99f904896ddcc73"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:27:49.248200Z","signature_b64":"VJEUFLbff6PABLQJo2wck7J28R8YsYURSA70hxluOwj4F1gEXI32CQdM4Lf6ul+QQHIX4yGHBwh0sgt/X3HYBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"287a297efaa435d3b53a35806be7486d5ce9241c2dd0b61d2c317a49ede192d9","last_reissued_at":"2026-07-05T08:27:49.247648Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:27:49.247648Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Grokfast: Accelerated Grokking by Amplifying Slow Gradients","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Bong Gyun Kang, Jaerin Lee, Kihoon Kim, Kyoung Mu Lee","submitted_at":"2024-05-30T16:35:30Z","abstract_excerpt":"One puzzling artifact in machine learning dubbed grokking is where delayed generalization is achieved tenfolds of iterations after near perfect overfitting to the training data. Focusing on the long delay itself on behalf of machine learning practitioners, our goal is to accelerate generalization of a model under grokking phenomenon. By regarding a series of gradients of a parameter over training iterations as a random signal over time, we can spectrally decompose the parameter trajectories under gradient descent into two components: the fast-varying, overfitting-yielding component and the slo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.20233","kind":"arxiv","version":2},"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/2405.20233/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":"2405.20233","created_at":"2026-07-05T08:27:49.247708+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.20233v2","created_at":"2026-07-05T08:27:49.247708+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.20233","created_at":"2026-07-05T08:27:49.247708+00:00"},{"alias_kind":"pith_short_12","alias_value":"FB5CS7X2UQ25","created_at":"2026-07-05T08:27:49.247708+00:00"},{"alias_kind":"pith_short_16","alias_value":"FB5CS7X2UQ25HNJ2","created_at":"2026-07-05T08:27:49.247708+00:00"},{"alias_kind":"pith_short_8","alias_value":"FB5CS7X2","created_at":"2026-07-05T08:27:49.247708+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":9,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.12966","citing_title":"Circuit Synchronization Precedes Generalization: A Causal Precursor to Grokking","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.32000","citing_title":"Radial Suppression Accelerates Algorithmic Generalization: A Geometric Analysis of Delayed Generalization","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2510.04930","citing_title":"Egalitarian Gradient Descent: A Simple Approach to Accelerated Grokking","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08464","citing_title":"The Geometric Structure of Models Learning Sparse Data","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12394","citing_title":"Detecting overfitting in Neural Networks during long-horizon grokking using Random Matrix Theory","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12394","citing_title":"Detecting overfitting in Neural Networks during long-horizon grokking using Random Matrix Theory","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08464","citing_title":"The Geometric Structure of Models Learning Sparse Data","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09724","citing_title":"Model Capacity Determines Grokking through Competing Memorisation and Generalisation Speeds","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20923","citing_title":"ILDR: Geometric Early Detection of Grokking","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FB5CS7X2UQ25HNJ2GWAGXZ2INV","json":"https://pith.science/pith/FB5CS7X2UQ25HNJ2GWAGXZ2INV.json","graph_json":"https://pith.science/api/pith-number/FB5CS7X2UQ25HNJ2GWAGXZ2INV/graph.json","events_json":"https://pith.science/api/pith-number/FB5CS7X2UQ25HNJ2GWAGXZ2INV/events.json","paper":"https://pith.science/paper/FB5CS7X2"},"agent_actions":{"view_html":"https://pith.science/pith/FB5CS7X2UQ25HNJ2GWAGXZ2INV","download_json":"https://pith.science/pith/FB5CS7X2UQ25HNJ2GWAGXZ2INV.json","view_paper":"https://pith.science/paper/FB5CS7X2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.20233&json=true","fetch_graph":"https://pith.science/api/pith-number/FB5CS7X2UQ25HNJ2GWAGXZ2INV/graph.json","fetch_events":"https://pith.science/api/pith-number/FB5CS7X2UQ25HNJ2GWAGXZ2INV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FB5CS7X2UQ25HNJ2GWAGXZ2INV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FB5CS7X2UQ25HNJ2GWAGXZ2INV/action/storage_attestation","attest_author":"https://pith.science/pith/FB5CS7X2UQ25HNJ2GWAGXZ2INV/action/author_attestation","sign_citation":"https://pith.science/pith/FB5CS7X2UQ25HNJ2GWAGXZ2INV/action/citation_signature","submit_replication":"https://pith.science/pith/FB5CS7X2UQ25HNJ2GWAGXZ2INV/action/replication_record"}},"created_at":"2026-07-05T08:27:49.247708+00:00","updated_at":"2026-07-05T08:27:49.247708+00:00"}