{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:JC54ULOC2QYNFIOWG7JDK5BJLZ","short_pith_number":"pith:JC54ULOC","canonical_record":{"source":{"id":"2605.17718","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-18T00:41:20Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"3d6dbf9d303aa6cc243e652f2e1946ff3b62a373824bc5c5037c7fb308bc4d4c","abstract_canon_sha256":"04dfd05f40b0f909587a10481e6fcc0d13e1e881075ad72eeca51348ce04fadd"},"schema_version":"1.0"},"canonical_sha256":"48bbca2dc2d430d2a1d637d23574295e7e44ffa4f6e31843a2f7ae6cfb809efb","source":{"kind":"arxiv","id":"2605.17718","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.17718","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"arxiv_version","alias_value":"2605.17718v1","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17718","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"pith_short_12","alias_value":"JC54ULOC2QYN","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"pith_short_16","alias_value":"JC54ULOC2QYNFIOW","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"pith_short_8","alias_value":"JC54ULOC","created_at":"2026-05-20T00:04:54Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:JC54ULOC2QYNFIOWG7JDK5BJLZ","target":"record","payload":{"canonical_record":{"source":{"id":"2605.17718","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-18T00:41:20Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"3d6dbf9d303aa6cc243e652f2e1946ff3b62a373824bc5c5037c7fb308bc4d4c","abstract_canon_sha256":"04dfd05f40b0f909587a10481e6fcc0d13e1e881075ad72eeca51348ce04fadd"},"schema_version":"1.0"},"canonical_sha256":"48bbca2dc2d430d2a1d637d23574295e7e44ffa4f6e31843a2f7ae6cfb809efb","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:04:54.554608Z","signature_b64":"t7ZMpsEcgA4R07Pr9iu/M1s5ClwQP1HeD6mPDx1UVLCrxwbU3rMuuRChsT0HRlEb7Pjx3bKyN4xKWgEVq2dlAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"48bbca2dc2d430d2a1d637d23574295e7e44ffa4f6e31843a2f7ae6cfb809efb","last_reissued_at":"2026-05-20T00:04:54.553700Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:04:54.553700Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.17718","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:04:54Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"9XBJMfyISe5OPH2KeX/HcoYieUc5WAiEv9m7e0RoHPTGK99qex5mRs7p4oX81zk3LoyxJY0EW0F5CyUUl6ISAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-24T17:17:10.251629Z"},"content_sha256":"aeb4422da15a1971e8305487473e9ce2328e7c41a91af405bbe0736c18baab87","schema_version":"1.0","event_id":"sha256:aeb4422da15a1971e8305487473e9ce2328e7c41a91af405bbe0736c18baab87"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:JC54ULOC2QYNFIOWG7JDK5BJLZ","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"How does feature learning reshape the function space?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"In high dimensions, one large gradient step on a two-layer network produces features whose distribution approximates a target-dependent spiked Gaussian covariance, inducing a data-adaptive kernel that reshapes the function space.","cross_cats":["cs.LG"],"primary_cat":"stat.ML","authors_text":"Bruno Loureiro, Fanghui Liu, Jo\\~ao Lobo, Long Tran-Than","submitted_at":"2026-05-18T00:41:20Z","abstract_excerpt":"Feature learning is widely regarded as the key mechanism distinguishing neural networks from fixed-kernel methods, yet its impact on the induced function space remains poorly understood. In this work, we precisely characterize how the function space spanned by the features of a two-layer neural network evolves during gradient descent training. We prove that, in the high-dimensional proportional regime, after a large gradient step the post-update feature distribution is well approximated by a target-dependent spiked Gaussian covariance. This induces a data-adaptive kernel that reshapes the func"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We prove that, in the high-dimensional proportional regime, after a large gradient step the post-update feature distribution is well approximated by a target-dependent spiked Gaussian covariance. This induces a data-adaptive kernel that reshapes the function space and modifies its spectral structure.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The analysis assumes the high-dimensional proportional regime (n, d → ∞ with n/d fixed) together with a sufficiently large gradient step size that allows the post-update feature distribution to be approximated by the spiked Gaussian form; if this regime or step-size condition fails to hold, the claimed approximation and resulting kernel reshaping do not necessarily follow.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"In the high-dimensional proportional regime, a large gradient step on a two-layer network induces a target-dependent spiked Gaussian covariance on the features, yielding a data-adaptive kernel that amplifies target-aligned eigenvalues and mixes leading eigenfunctions.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"In high dimensions, one large gradient step on a two-layer network produces features whose distribution approximates a target-dependent spiked Gaussian covariance, inducing a data-adaptive kernel that reshapes the function space.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e57d426033df0e0883378f14b079a49c3eb394a76258dba680ce88c02c8df993"},"source":{"id":"2605.17718","kind":"arxiv","version":1},"verdict":{"id":"5516f7ea-633a-480e-b5fa-f4012773bed4","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-19T22:22:46.075133Z","strongest_claim":"We prove that, in the high-dimensional proportional regime, after a large gradient step the post-update feature distribution is well approximated by a target-dependent spiked Gaussian covariance. This induces a data-adaptive kernel that reshapes the function space and modifies its spectral structure.","one_line_summary":"In the high-dimensional proportional regime, a large gradient step on a two-layer network induces a target-dependent spiked Gaussian covariance on the features, yielding a data-adaptive kernel that amplifies target-aligned eigenvalues and mixes leading eigenfunctions.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The analysis assumes the high-dimensional proportional regime (n, d → ∞ with n/d fixed) together with a sufficiently large gradient step size that allows the post-update feature distribution to be approximated by the spiked Gaussian form; if this regime or step-size condition fails to hold, the claimed approximation and resulting kernel reshaping do not necessarily follow.","pith_extraction_headline":"In high dimensions, one large gradient step on a two-layer network produces features whose distribution approximates a target-dependent spiked Gaussian covariance, inducing a data-adaptive kernel that reshapes the function space."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.17718/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T22:31:19.395469Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T22:31:02.068336Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"shingle_duplication","ran_at":"2026-05-19T21:49:43.469108Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"citation_quote_validity","ran_at":"2026-05-19T21:49:43.298583Z","status":"skipped","version":"0.1.0","findings_count":0},{"name":"ai_meta_artifact","ran_at":"2026-05-19T21:33:23.502463Z","status":"skipped","version":"1.0.0","findings_count":0},{"name":"cited_work_retraction","ran_at":"2026-05-19T21:21:57.725628Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"claim_evidence","ran_at":"2026-05-19T21:21:57.405773Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"82d51dcc2179da61e7e06d6fedc55ffc21092a56b10dd788a168a94f0659f04b"},"references":{"count":300,"sample":[{"doi":"10.1038/s41586-020-2649-2","year":2020,"title":"Charles R. Harris and K. Jarrod Millman and St. Array programming with. 2020 , month = sep, journal =. doi:10.1038/s41586-020-2649-2 , publisher =","work_id":"b05b154d-0381-4d1b-911f-1b35eb7a6768","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and","work_id":"85fe8ea2-dfba-4883-bbba-2341408e0f7b","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Pedregosa, F. and Varoquaux, G. and Gramfort, A. and Michel, V. and Thirion, B. and Grisel, O. and Blondel, M. and Prettenhofer, P. and Weiss, R. and Dubourg, V. and Vanderplas, J. and Passos, A. and ","work_id":"e9bde9af-fdb4-4d1b-8606-58708d1b8200","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"PyTorch: An Imperative Style, High-Performance Deep Learning Library , url =","work_id":"ceff234f-9a58-4c0a-bcce-766fb47fc4bb","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":null,"title":"Journal of Machine Learning Research , year =","work_id":"bc3f0c2f-7dae-4bf7-8aae-1609f1a5ca20","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":300,"snapshot_sha256":"bb64dd9a998d7736a77c10b20f9ecad1b7a8cba110167d2edc266b1d7ef0ba52","internal_anchors":2},"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"},"verdict_id":"5516f7ea-633a-480e-b5fa-f4012773bed4"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:04:54Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"r8ZzI0MQEAiXj6ptB1SbyVh7gXWKg91xMV4y6yEGMHVYwgqx5Q+TixgWRsfIXeOvoKzjPyBAnM6ugff7u7cvDA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-05-24T17:17:10.253292Z"},"content_sha256":"a9a99f7463cf6ac048bd7a38b6864b692cb21104d774bffbc88c24a31e2dfb26","schema_version":"1.0","event_id":"sha256:a9a99f7463cf6ac048bd7a38b6864b692cb21104d774bffbc88c24a31e2dfb26"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/JC54ULOC2QYNFIOWG7JDK5BJLZ/bundle.json","state_url":"https://pith.science/pith/JC54ULOC2QYNFIOWG7JDK5BJLZ/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/JC54ULOC2QYNFIOWG7JDK5BJLZ/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-05-24T17:17:10Z","links":{"resolver":"https://pith.science/pith/JC54ULOC2QYNFIOWG7JDK5BJLZ","bundle":"https://pith.science/pith/JC54ULOC2QYNFIOWG7JDK5BJLZ/bundle.json","state":"https://pith.science/pith/JC54ULOC2QYNFIOWG7JDK5BJLZ/state.json","well_known_bundle":"https://pith.science/.well-known/pith/JC54ULOC2QYNFIOWG7JDK5BJLZ/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:JC54ULOC2QYNFIOWG7JDK5BJLZ","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":"04dfd05f40b0f909587a10481e6fcc0d13e1e881075ad72eeca51348ce04fadd","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-18T00:41:20Z","title_canon_sha256":"3d6dbf9d303aa6cc243e652f2e1946ff3b62a373824bc5c5037c7fb308bc4d4c"},"schema_version":"1.0","source":{"id":"2605.17718","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.17718","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"arxiv_version","alias_value":"2605.17718v1","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.17718","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"pith_short_12","alias_value":"JC54ULOC2QYN","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"pith_short_16","alias_value":"JC54ULOC2QYNFIOW","created_at":"2026-05-20T00:04:54Z"},{"alias_kind":"pith_short_8","alias_value":"JC54ULOC","created_at":"2026-05-20T00:04:54Z"}],"graph_snapshots":[{"event_id":"sha256:a9a99f7463cf6ac048bd7a38b6864b692cb21104d774bffbc88c24a31e2dfb26","target":"graph","created_at":"2026-05-20T00:04:54Z","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":"We prove that, in the high-dimensional proportional regime, after a large gradient step the post-update feature distribution is well approximated by a target-dependent spiked Gaussian covariance. This induces a data-adaptive kernel that reshapes the function space and modifies its spectral structure."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The analysis assumes the high-dimensional proportional regime (n, d → ∞ with n/d fixed) together with a sufficiently large gradient step size that allows the post-update feature distribution to be approximated by the spiked Gaussian form; if this regime or step-size condition fails to hold, the claimed approximation and resulting kernel reshaping do not necessarily follow."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"In the high-dimensional proportional regime, a large gradient step on a two-layer network induces a target-dependent spiked Gaussian covariance on the features, yielding a data-adaptive kernel that amplifies target-aligned eigenvalues and mixes leading eigenfunctions."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"In high dimensions, one large gradient step on a two-layer network produces features whose distribution approximates a target-dependent spiked Gaussian covariance, inducing a data-adaptive kernel that reshapes the function space."}],"snapshot_sha256":"e57d426033df0e0883378f14b079a49c3eb394a76258dba680ce88c02c8df993"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T22:31:19.395469Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T22:31:02.068336Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"shingle_duplication","ran_at":"2026-05-19T21:49:43.469108Z","status":"skipped","version":"0.1.0"},{"findings_count":0,"name":"citation_quote_validity","ran_at":"2026-05-19T21:49:43.298583Z","status":"skipped","version":"0.1.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-19T21:33:23.502463Z","status":"skipped","version":"1.0.0"},{"findings_count":0,"name":"cited_work_retraction","ran_at":"2026-05-19T21:21:57.725628Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-19T21:21:57.405773Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.17718/integrity.json","findings":[],"snapshot_sha256":"82d51dcc2179da61e7e06d6fedc55ffc21092a56b10dd788a168a94f0659f04b","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Feature learning is widely regarded as the key mechanism distinguishing neural networks from fixed-kernel methods, yet its impact on the induced function space remains poorly understood. In this work, we precisely characterize how the function space spanned by the features of a two-layer neural network evolves during gradient descent training. We prove that, in the high-dimensional proportional regime, after a large gradient step the post-update feature distribution is well approximated by a target-dependent spiked Gaussian covariance. This induces a data-adaptive kernel that reshapes the func","authors_text":"Bruno Loureiro, Fanghui Liu, Jo\\~ao Lobo, Long Tran-Than","cross_cats":["cs.LG"],"headline":"In high dimensions, one large gradient step on a two-layer network produces features whose distribution approximates a target-dependent spiked Gaussian covariance, inducing a data-adaptive kernel that reshapes the function space.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-18T00:41:20Z","title":"How does feature learning reshape the function space?"},"references":{"count":300,"internal_anchors":2,"resolved_work":300,"sample":[{"cited_arxiv_id":"","doi":"10.1038/s41586-020-2649-2","is_internal_anchor":false,"ref_index":1,"title":"Charles R. Harris and K. Jarrod Millman and St. Array programming with. 2020 , month = sep, journal =. doi:10.1038/s41586-020-2649-2 , publisher =","work_id":"b05b154d-0381-4d1b-911f-1b35eb7a6768","year":2020},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"and Haberland, Matt and Reddy, Tyler and Cournapeau, David and Burovski, Evgeni and Peterson, Pearu and Weckesser, Warren and Bright, Jonathan and","work_id":"85fe8ea2-dfba-4883-bbba-2341408e0f7b","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Pedregosa, F. and Varoquaux, G. and Gramfort, A. and Michel, V. and Thirion, B. and Grisel, O. and Blondel, M. and Prettenhofer, P. and Weiss, R. and Dubourg, V. and Vanderplas, J. and Passos, A. and ","work_id":"e9bde9af-fdb4-4d1b-8606-58708d1b8200","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"PyTorch: An Imperative Style, High-Performance Deep Learning Library , url =","work_id":"ceff234f-9a58-4c0a-bcce-766fb47fc4bb","year":null},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"Journal of Machine Learning Research , year =","work_id":"bc3f0c2f-7dae-4bf7-8aae-1609f1a5ca20","year":null}],"snapshot_sha256":"bb64dd9a998d7736a77c10b20f9ecad1b7a8cba110167d2edc266b1d7ef0ba52"},"source":{"id":"2605.17718","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-19T22:22:46.075133Z","id":"5516f7ea-633a-480e-b5fa-f4012773bed4","model_set":{"reader":"grok-4.3"},"one_line_summary":"In the high-dimensional proportional regime, a large gradient step on a two-layer network induces a target-dependent spiked Gaussian covariance on the features, yielding a data-adaptive kernel that amplifies target-aligned eigenvalues and mixes leading eigenfunctions.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"In high dimensions, one large gradient step on a two-layer network produces features whose distribution approximates a target-dependent spiked Gaussian covariance, inducing a data-adaptive kernel that reshapes the function space.","strongest_claim":"We prove that, in the high-dimensional proportional regime, after a large gradient step the post-update feature distribution is well approximated by a target-dependent spiked Gaussian covariance. This induces a data-adaptive kernel that reshapes the function space and modifies its spectral structure.","weakest_assumption":"The analysis assumes the high-dimensional proportional regime (n, d → ∞ with n/d fixed) together with a sufficiently large gradient step size that allows the post-update feature distribution to be approximated by the spiked Gaussian form; if this regime or step-size condition fails to hold, the claimed approximation and resulting kernel reshaping do not necessarily follow."}},"verdict_id":"5516f7ea-633a-480e-b5fa-f4012773bed4"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:aeb4422da15a1971e8305487473e9ce2328e7c41a91af405bbe0736c18baab87","target":"record","created_at":"2026-05-20T00:04:54Z","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":"04dfd05f40b0f909587a10481e6fcc0d13e1e881075ad72eeca51348ce04fadd","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"stat.ML","submitted_at":"2026-05-18T00:41:20Z","title_canon_sha256":"3d6dbf9d303aa6cc243e652f2e1946ff3b62a373824bc5c5037c7fb308bc4d4c"},"schema_version":"1.0","source":{"id":"2605.17718","kind":"arxiv","version":1}},"canonical_sha256":"48bbca2dc2d430d2a1d637d23574295e7e44ffa4f6e31843a2f7ae6cfb809efb","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"48bbca2dc2d430d2a1d637d23574295e7e44ffa4f6e31843a2f7ae6cfb809efb","first_computed_at":"2026-05-20T00:04:54.553700Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:04:54.553700Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"t7ZMpsEcgA4R07Pr9iu/M1s5ClwQP1HeD6mPDx1UVLCrxwbU3rMuuRChsT0HRlEb7Pjx3bKyN4xKWgEVq2dlAQ==","signature_status":"signed_v1","signed_at":"2026-05-20T00:04:54.554608Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.17718","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:aeb4422da15a1971e8305487473e9ce2328e7c41a91af405bbe0736c18baab87","sha256:a9a99f7463cf6ac048bd7a38b6864b692cb21104d774bffbc88c24a31e2dfb26"],"state_sha256":"ad2b2ed4eb3ad97e1172dd5c0bdf0c2b531fb87b4f27d04495631215992890ef"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"wzWZBLsWVsMQ2/gG32HkJwPDOhLA25Fn29mqR4pU9QyqcQ1eoCkaIK1Q59Ob3FuG9ON78FIVylZSckKscmC5Cg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-05-24T17:17:10.258683Z","bundle_sha256":"86ffa1a60dfdbc938781fb24dfaba412011552eee2c261996f15ac9665f0ef7d"}}