{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:ON25ECWU7YVOFZ7KHJWRFYOHJS","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"91734d4062d34a7c318692774f524346ae498c97a0a9ae90bc1e897a3636e805","cross_cats_sorted":["cs.AI","math.PR","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-12-24T09:39:04Z","title_canon_sha256":"f681a678a515c588697b44471af8b50a099a038d207515149b5a0ea0d3534034"},"schema_version":"1.0","source":{"id":"2512.21075","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2512.21075","created_at":"2026-05-18T03:09:32Z"},{"alias_kind":"arxiv_version","alias_value":"2512.21075v2","created_at":"2026-05-18T03:09:32Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2512.21075","created_at":"2026-05-18T03:09:32Z"},{"alias_kind":"pith_short_12","alias_value":"ON25ECWU7YVO","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"ON25ECWU7YVOFZ7K","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"ON25ECWU","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:84f5b1a48f1f7fac5da34d2785cc3be2568f07c31f952cf5eed5604064fc2991","target":"graph","created_at":"2026-05-18T03:09:32Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Under nondegeneracy assumptions, we prove that the finite-network training dynamics converge to its NFD limit with an O(L^{-1}) depth-discretization error, while the reused-weight coupling term has a faster O(L^{-2}) decay."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The nondegeneracy assumptions on the feature-gradient covariance structure generated during training, which are required to ensure the SDE limit exists and that the coupling remains higher-order in depth under depth-μP scaling."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Under depth-μP scaling, the reused-weight forward-backward coupling in one-layer ResNets vanishes at O(L^{-2}), enabling convergence to a decoupled Neural Feature Dynamics SDE limit with O(L^{-1}) discretization error."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Finite ResNet training dynamics converge to a decoupled Neural Feature Dynamics limit with O(L^{-1}) error under depth-μP scaling."}],"snapshot_sha256":"87b8d87b67f7500719b5b9f349a1b632e89dc4d4b0f9b67dd2b0300ba00d78f5"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"47532bf44977a180848007fd5de320cc7c0c3e8dcc6ba27e37503faae83f9a12"},"paper":{"abstract_excerpt":"Deep neural networks have achieved remarkable success in practice, yet a mechanistic understanding of how features evolve during training remains incomplete, especially in the large-depth limit. For ResNets under depth-$\\mu$P scaling, prior work treats the layer index $\\ell$ as a continuous time $t_\\ell = \\ell/L$, yielding SDE descriptions of the training dynamics. A key unresolved issue is that backpropagation reuses each forward weight matrix $W_\\ell$ through its transpose $W_\\ell^\\top$, creating correlations between forward features and backward gradients whose behavior and role in feature ","authors_text":"Ruoyu Wu, Tianxiang Gao, Zihan Yao","cross_cats":["cs.AI","math.PR","stat.ML"],"headline":"Finite ResNet training dynamics converge to a decoupled Neural Feature Dynamics limit with O(L^{-1}) error under depth-μP scaling.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-12-24T09:39:04Z","title":"Feature Learning Dynamics in Infinite-Depth Neural Networks"},"references":{"count":4,"internal_anchors":0,"resolved_work":4,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"write newline","work_id":"8e5fda61-e601-4df4-8204-015bee341570","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"\\@ifxundefined[1] #1\\@undefined \\@firstoftwo \\@secondoftwo \\@ifnum[1] #1 \\@firstoftwo \\@secondoftwo \\@ifx[1] #1 \\@firstoftwo \\@secondoftwo [2] @ #1 \\@temptokena #2 #1 @ \\@temptokena \\@ifclassloaded ag","work_id":"b058608d-98d0-4821-a4ae-403d2b7cd411","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"\\@lbibitem[] @bibitem@first@sw\\@secondoftwo \\@lbibitem[#1]#2 \\@extra@b@citeb \\@ifundefined br@#2\\@extra@b@citeb \\@namedef br@#2 \\@nameuse br@#2\\@extra@b@citeb \\@ifundefined b@#2\\@extra@b@citeb @num @p","work_id":"ea79bfb8-d434-45e9-8607-416d3839ec5c","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"@open @close @open @close and [1] URL: #1 \\@ifundefined chapter * \\@mkboth \\@ifxundefined @sectionbib * \\@mkboth * \\@mkboth\\@gobbletwo \\@ifclassloaded amsart * \\@ifclassloaded amsbook * \\@ifxundefined","work_id":"d051d6aa-efca-49eb-a66d-8ea01bf21294","year":null}],"snapshot_sha256":"8cd3c2443fe68f5b0b489cafcb4da447febf1db446e7328adbfe038ec9ee0c09"},"source":{"id":"2512.21075","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-16T20:13:31.612520Z","id":"a4c93aa3-e0ce-4656-8417-7d9fe1a74f46","model_set":{"reader":"grok-4.3"},"one_line_summary":"Under depth-μP scaling, the reused-weight forward-backward coupling in one-layer ResNets vanishes at O(L^{-2}), enabling convergence to a decoupled Neural Feature Dynamics SDE limit with O(L^{-1}) discretization error.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Finite ResNet training dynamics converge to a decoupled Neural Feature Dynamics limit with O(L^{-1}) error under depth-μP scaling.","strongest_claim":"Under nondegeneracy assumptions, we prove that the finite-network training dynamics converge to its NFD limit with an O(L^{-1}) depth-discretization error, while the reused-weight coupling term has a faster O(L^{-2}) decay.","weakest_assumption":"The nondegeneracy assumptions on the feature-gradient covariance structure generated during training, which are required to ensure the SDE limit exists and that the coupling remains higher-order in depth under depth-μP scaling."}},"verdict_id":"a4c93aa3-e0ce-4656-8417-7d9fe1a74f46"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:173633dbb0051a96ebe0b3bd567e5478daa923bdf8abe34046d1a2f0d77cc066","target":"record","created_at":"2026-05-18T03:09:32Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"91734d4062d34a7c318692774f524346ae498c97a0a9ae90bc1e897a3636e805","cross_cats_sorted":["cs.AI","math.PR","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-12-24T09:39:04Z","title_canon_sha256":"f681a678a515c588697b44471af8b50a099a038d207515149b5a0ea0d3534034"},"schema_version":"1.0","source":{"id":"2512.21075","kind":"arxiv","version":2}},"canonical_sha256":"7375d20ad4fe2ae2e7ea3a6d12e1c74c9ce0aeb71e2cad3b977bbf207b8942c1","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"7375d20ad4fe2ae2e7ea3a6d12e1c74c9ce0aeb71e2cad3b977bbf207b8942c1","first_computed_at":"2026-05-18T03:09:32.417066Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T03:09:32.417066Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"smXP2zTB3L4AWZ8Um5iBUgTWVRLWltc/K1ghk+f3wd4pAdJEZ7U0FRDIRTdUh9xEW4J2HWY1ED8WrE2okKoGBg==","signature_status":"signed_v1","signed_at":"2026-05-18T03:09:32.417954Z","signed_message":"canonical_sha256_bytes"},"source_id":"2512.21075","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:173633dbb0051a96ebe0b3bd567e5478daa923bdf8abe34046d1a2f0d77cc066","sha256:84f5b1a48f1f7fac5da34d2785cc3be2568f07c31f952cf5eed5604064fc2991"],"state_sha256":"04d3f22cd89337be5c0a5e524199ebff7fde72dbec9bc039b6300d56e3ebc22d"}