{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:U7SXCBA4Y26GMBABEMZNJAKZBC","short_pith_number":"pith:U7SXCBA4","schema_version":"1.0","canonical_sha256":"a7e571041cc6bc6604012332d4815908bf18c89b6f42bc0108cda9a1730a2852","source":{"kind":"arxiv","id":"2505.13900","version":1},"attestation_state":"computed","paper":{"title":"New Evidence of the Two-Phase Learning Dynamics of Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Junchi Yan, Mahito Sugiyama, Yongyi Yang, Zhanpeng Zhou","submitted_at":"2025-05-20T04:03:52Z","abstract_excerpt":"Understanding how deep neural networks learn remains a fundamental challenge in modern machine learning. A growing body of evidence suggests that training dynamics undergo a distinct phase transition, yet our understanding of this transition is still incomplete. In this paper, we introduce an interval-wise perspective that compares network states across a time window, revealing two new phenomena that illuminate the two-phase nature of deep learning. i) \\textbf{The Chaos Effect.} By injecting an imperceptibly small parameter perturbation at various stages, we show that the response of the netwo"},"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":"2505.13900","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-20T04:03:52Z","cross_cats_sorted":[],"title_canon_sha256":"5b85ccd2e740fbb0881961445bbb05a2a9a6166802cc14fc813f60c926f1e70a","abstract_canon_sha256":"a707e103449b6654e3867edc5af0b65a9ad22a5033bb763ac742b7bb239c1527"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:05:52.626020Z","signature_b64":"NV4Yemr2IM3U/OJaUMojj4VbLFdgdTWJ9cl/85yKtXno/8aKNIT6h2u5EzE+s9SNcB16aqdrY9qJbEYfHB8zDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a7e571041cc6bc6604012332d4815908bf18c89b6f42bc0108cda9a1730a2852","last_reissued_at":"2026-07-05T11:05:52.625588Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:05:52.625588Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"New Evidence of the Two-Phase Learning Dynamics of Neural Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Junchi Yan, Mahito Sugiyama, Yongyi Yang, Zhanpeng Zhou","submitted_at":"2025-05-20T04:03:52Z","abstract_excerpt":"Understanding how deep neural networks learn remains a fundamental challenge in modern machine learning. A growing body of evidence suggests that training dynamics undergo a distinct phase transition, yet our understanding of this transition is still incomplete. In this paper, we introduce an interval-wise perspective that compares network states across a time window, revealing two new phenomena that illuminate the two-phase nature of deep learning. i) \\textbf{The Chaos Effect.} By injecting an imperceptibly small parameter perturbation at various stages, we show that the response of the netwo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.13900","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/2505.13900/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":"2505.13900","created_at":"2026-07-05T11:05:52.625645+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.13900v1","created_at":"2026-07-05T11:05:52.625645+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.13900","created_at":"2026-07-05T11:05:52.625645+00:00"},{"alias_kind":"pith_short_12","alias_value":"U7SXCBA4Y26G","created_at":"2026-07-05T11:05:52.625645+00:00"},{"alias_kind":"pith_short_16","alias_value":"U7SXCBA4Y26GMBAB","created_at":"2026-07-05T11:05:52.625645+00:00"},{"alias_kind":"pith_short_8","alias_value":"U7SXCBA4","created_at":"2026-07-05T11:05:52.625645+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/U7SXCBA4Y26GMBABEMZNJAKZBC","json":"https://pith.science/pith/U7SXCBA4Y26GMBABEMZNJAKZBC.json","graph_json":"https://pith.science/api/pith-number/U7SXCBA4Y26GMBABEMZNJAKZBC/graph.json","events_json":"https://pith.science/api/pith-number/U7SXCBA4Y26GMBABEMZNJAKZBC/events.json","paper":"https://pith.science/paper/U7SXCBA4"},"agent_actions":{"view_html":"https://pith.science/pith/U7SXCBA4Y26GMBABEMZNJAKZBC","download_json":"https://pith.science/pith/U7SXCBA4Y26GMBABEMZNJAKZBC.json","view_paper":"https://pith.science/paper/U7SXCBA4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.13900&json=true","fetch_graph":"https://pith.science/api/pith-number/U7SXCBA4Y26GMBABEMZNJAKZBC/graph.json","fetch_events":"https://pith.science/api/pith-number/U7SXCBA4Y26GMBABEMZNJAKZBC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/U7SXCBA4Y26GMBABEMZNJAKZBC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/U7SXCBA4Y26GMBABEMZNJAKZBC/action/storage_attestation","attest_author":"https://pith.science/pith/U7SXCBA4Y26GMBABEMZNJAKZBC/action/author_attestation","sign_citation":"https://pith.science/pith/U7SXCBA4Y26GMBABEMZNJAKZBC/action/citation_signature","submit_replication":"https://pith.science/pith/U7SXCBA4Y26GMBABEMZNJAKZBC/action/replication_record"}},"created_at":"2026-07-05T11:05:52.625645+00:00","updated_at":"2026-07-05T11:05:52.625645+00:00"}