{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:CTUMD4333JF5CCIBXUUSOWENFW","short_pith_number":"pith:CTUMD433","schema_version":"1.0","canonical_sha256":"14e8c1f37bda4bd10901bd2927588d2daa1793aa0268f18314ebd36d69a2b903","source":{"kind":"arxiv","id":"2402.15175","version":2},"attestation_state":"computed","paper":{"title":"Unified View of Grokking, Double Descent and Emergent Abilities: A Perspective from Circuits Competition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Maosong Sun, Shengding Hu, Xu Han, Yufei Huang, Zhiyuan Liu","submitted_at":"2024-02-23T08:14:36Z","abstract_excerpt":"Recent studies have uncovered intriguing phenomena in deep learning, such as grokking, double descent, and emergent abilities in large language models, which challenge human intuition and are crucial for a deeper understanding of neural models. In this paper, we present a comprehensive framework that provides a unified view of these three phenomena, focusing on the competition between memorization and generalization circuits. This approach, initially employed to explain grokking, is extended in our work to encompass a wider range of model sizes and training data volumes. Our framework delineat"},"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":"2402.15175","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-02-23T08:14:36Z","cross_cats_sorted":[],"title_canon_sha256":"a1a1ec83acd8df66f07615ee5fb23cee1e47d2b3c9fde1128d82a55ab131a648","abstract_canon_sha256":"bbfb835eb745137b8e961b5783bc959f025989268bfe8b34514100c78803733b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:49:18.706880Z","signature_b64":"redfGa1+1V3Jk5m2ZjSRS8J6txLt1kI0rPNFNa8OhmTB4wBEos1095+z2/EnvaC9SLc18scJ062TgqHxkj7/CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"14e8c1f37bda4bd10901bd2927588d2daa1793aa0268f18314ebd36d69a2b903","last_reissued_at":"2026-07-05T07:49:18.706358Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:49:18.706358Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unified View of Grokking, Double Descent and Emergent Abilities: A Perspective from Circuits Competition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Maosong Sun, Shengding Hu, Xu Han, Yufei Huang, Zhiyuan Liu","submitted_at":"2024-02-23T08:14:36Z","abstract_excerpt":"Recent studies have uncovered intriguing phenomena in deep learning, such as grokking, double descent, and emergent abilities in large language models, which challenge human intuition and are crucial for a deeper understanding of neural models. In this paper, we present a comprehensive framework that provides a unified view of these three phenomena, focusing on the competition between memorization and generalization circuits. This approach, initially employed to explain grokking, is extended in our work to encompass a wider range of model sizes and training data volumes. Our framework delineat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.15175","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/2402.15175/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":"2402.15175","created_at":"2026-07-05T07:49:18.706431+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.15175v2","created_at":"2026-07-05T07:49:18.706431+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.15175","created_at":"2026-07-05T07:49:18.706431+00:00"},{"alias_kind":"pith_short_12","alias_value":"CTUMD4333JF5","created_at":"2026-07-05T07:49:18.706431+00:00"},{"alias_kind":"pith_short_16","alias_value":"CTUMD4333JF5CCIB","created_at":"2026-07-05T07:49:18.706431+00:00"},{"alias_kind":"pith_short_8","alias_value":"CTUMD433","created_at":"2026-07-05T07:49:18.706431+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.14659","citing_title":"Slower Generalization, Faster Memorization: A Sweet Spot in Algorithmic Learning","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18022","citing_title":"Unveiling Memorization-Generalization Coexistence: A Case Study on Arithmetic Tasks with Label Noise","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.09724","citing_title":"Model Capacity Determines Grokking through Competing Memorisation and Generalisation Speeds","ref_index":33,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CTUMD4333JF5CCIBXUUSOWENFW","json":"https://pith.science/pith/CTUMD4333JF5CCIBXUUSOWENFW.json","graph_json":"https://pith.science/api/pith-number/CTUMD4333JF5CCIBXUUSOWENFW/graph.json","events_json":"https://pith.science/api/pith-number/CTUMD4333JF5CCIBXUUSOWENFW/events.json","paper":"https://pith.science/paper/CTUMD433"},"agent_actions":{"view_html":"https://pith.science/pith/CTUMD4333JF5CCIBXUUSOWENFW","download_json":"https://pith.science/pith/CTUMD4333JF5CCIBXUUSOWENFW.json","view_paper":"https://pith.science/paper/CTUMD433","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.15175&json=true","fetch_graph":"https://pith.science/api/pith-number/CTUMD4333JF5CCIBXUUSOWENFW/graph.json","fetch_events":"https://pith.science/api/pith-number/CTUMD4333JF5CCIBXUUSOWENFW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CTUMD4333JF5CCIBXUUSOWENFW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CTUMD4333JF5CCIBXUUSOWENFW/action/storage_attestation","attest_author":"https://pith.science/pith/CTUMD4333JF5CCIBXUUSOWENFW/action/author_attestation","sign_citation":"https://pith.science/pith/CTUMD4333JF5CCIBXUUSOWENFW/action/citation_signature","submit_replication":"https://pith.science/pith/CTUMD4333JF5CCIBXUUSOWENFW/action/replication_record"}},"created_at":"2026-07-05T07:49:18.706431+00:00","updated_at":"2026-07-05T07:49:18.706431+00:00"}