{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:3VKFXRQM6RXARY7J43WUMGYQRO","short_pith_number":"pith:3VKFXRQM","canonical_record":{"source":{"id":"2605.01288","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-02T06:55:15Z","cross_cats_sorted":["cond-mat.dis-nn","stat.ML"],"title_canon_sha256":"52163603e99629cbdd0bb6cb17146be35167df9f9963c970d5f3e49cf163a903","abstract_canon_sha256":"8635d78054aeba55bcc86b93793aeb1cb6e41a0c5a13564e20ba2274e366a0c6"},"schema_version":"1.0"},"canonical_sha256":"dd545bc60cf46e08e3e9e6ed461b108b829581fa3e9cacc5ccac9f69c012f4bf","source":{"kind":"arxiv","id":"2605.01288","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.01288","created_at":"2026-06-24T01:15:03Z"},{"alias_kind":"arxiv_version","alias_value":"2605.01288v3","created_at":"2026-06-24T01:15:03Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.01288","created_at":"2026-06-24T01:15:03Z"},{"alias_kind":"pith_short_12","alias_value":"3VKFXRQM6RXA","created_at":"2026-06-24T01:15:03Z"},{"alias_kind":"pith_short_16","alias_value":"3VKFXRQM6RXARY7J","created_at":"2026-06-24T01:15:03Z"},{"alias_kind":"pith_short_8","alias_value":"3VKFXRQM","created_at":"2026-06-24T01:15:03Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:3VKFXRQM6RXARY7J43WUMGYQRO","target":"record","payload":{"canonical_record":{"source":{"id":"2605.01288","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-02T06:55:15Z","cross_cats_sorted":["cond-mat.dis-nn","stat.ML"],"title_canon_sha256":"52163603e99629cbdd0bb6cb17146be35167df9f9963c970d5f3e49cf163a903","abstract_canon_sha256":"8635d78054aeba55bcc86b93793aeb1cb6e41a0c5a13564e20ba2274e366a0c6"},"schema_version":"1.0"},"canonical_sha256":"dd545bc60cf46e08e3e9e6ed461b108b829581fa3e9cacc5ccac9f69c012f4bf","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-24T01:15:03.269368Z","signature_b64":"jQa4d9G+psq9z21BNP9BthOQ/bBctVqlvYM7Te89TWjWOLXMaVYb43cjIy9UQ8SnFheUmDrJF9y/DPgFBZ8oAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dd545bc60cf46e08e3e9e6ed461b108b829581fa3e9cacc5ccac9f69c012f4bf","last_reissued_at":"2026-06-24T01:15:03.268859Z","signature_status":"signed_v1","first_computed_at":"2026-06-24T01:15:03.268859Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.01288","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-24T01:15:03Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"UM7a6JDJuiNcn/jxzj3VF9FO2mgY2zz1YwQQowzgMb3cZrIQry3siAocGq0InftTAA4+gY+DT6DZDRnEa/QkBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T17:27:45.301945Z"},"content_sha256":"c3c64c40e7dce5de1e9a1040c9b79ed12828dbdc3ff98caea5470a9e97661bed","schema_version":"1.0","event_id":"sha256:c3c64c40e7dce5de1e9a1040c9b79ed12828dbdc3ff98caea5470a9e97661bed"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:3VKFXRQM6RXARY7J43WUMGYQRO","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"A Theory of Saddle Escape in Deep Nonlinear Networks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"An exact identity on Frobenius norm imbalances in layer weights reduces deep nonlinear training to a scalar ODE whose escape time scales as ε to the power of minus (r minus 2), where r is the number of bottleneck layers.","cross_cats":["cond-mat.dis-nn","stat.ML"],"primary_cat":"cs.LG","authors_text":"Divit Rawal, Michael R. DeWeese","submitted_at":"2026-05-02T06:55:15Z","abstract_excerpt":"In deep networks with small initialization, training exhibits long plateaus separated by sharp feature-acquisition transitions. Whereas shallow nonlinear networks and deep linear networks are well studied, extending these analyses to deep nonlinear networks remains challenging. We derive an exact identity for the imbalance of Frobenius norms of layer weight matrices that holds for any smooth activation and any differentiable loss and use this to classify activation functions into four universality classes. On the permutation-symmetric submanifold, the identity combines with an approximate bala"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We derive an exact identity for the imbalance of Frobenius norms of layer weight matrices that holds for any smooth activation and any differentiable loss [...] giving a critical-depth escape time law τ★ = Θ(ε^{-(r-2)}) governed by the number r of layers at the bottleneck scale rather than the total depth L.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The reduction to a scalar ODE relies on an approximate balance law on the permutation-symmetric submanifold whose accuracy and range of validity are not derived from first principles but stated as holding approximately.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"An exact norm-imbalance identity classifies activations into four classes and reduces deep nonlinear training flow to a scalar ODE that predicts saddle escape time scaling as ε to the power of minus (r-2) for r bottleneck layers.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"An exact identity on Frobenius norm imbalances in layer weights reduces deep nonlinear training to a scalar ODE whose escape time scales as ε to the power of minus (r minus 2), where r is the number of bottleneck layers.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f6bfc38509f467af70de9264d04f9b884fdae7fcd6580e2a244fabe181f13984"},"source":{"id":"2605.01288","kind":"arxiv","version":3},"verdict":{"id":"79015d10-f517-4bfd-958a-c773aae68005","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-11T02:19:07.595331Z","strongest_claim":"We derive an exact identity for the imbalance of Frobenius norms of layer weight matrices that holds for any smooth activation and any differentiable loss [...] giving a critical-depth escape time law τ★ = Θ(ε^{-(r-2)}) governed by the number r of layers at the bottleneck scale rather than the total depth L.","one_line_summary":"An exact norm-imbalance identity classifies activations into four classes and reduces deep nonlinear training flow to a scalar ODE that predicts saddle escape time scaling as ε to the power of minus (r-2) for r bottleneck layers.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The reduction to a scalar ODE relies on an approximate balance law on the permutation-symmetric submanifold whose accuracy and range of validity are not derived from first principles but stated as holding approximately.","pith_extraction_headline":"An exact identity on Frobenius norm imbalances in layer weights reduces deep nonlinear training to a scalar ODE whose escape time scales as ε to the power of minus (r minus 2), where r is the number of bottleneck layers."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.01288/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T18:35:43.254592Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T17:26:57.271652Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"952980ec58cd8c73efa880119863bdc6b10d930db90d3c8491886652e2a991f4"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"37a72047ffc37f0b237735c20891583088a0a2d5e11361d6c33f9df2470661db"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"79015d10-f517-4bfd-958a-c773aae68005"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-24T01:15:03Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"QRT+y1sM9O+DDXUufi9v37cDv/ifTDMXu6oDO3hM93jNX4d09Z+w/VSgveTlj5xrj3pbdQ/oPFcbHJ54GfEnBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T17:27:45.302832Z"},"content_sha256":"559760e5bc3265b033a9495c06de415d513b608fee65c87974567adb163d349c","schema_version":"1.0","event_id":"sha256:559760e5bc3265b033a9495c06de415d513b608fee65c87974567adb163d349c"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/3VKFXRQM6RXARY7J43WUMGYQRO/bundle.json","state_url":"https://pith.science/pith/3VKFXRQM6RXARY7J43WUMGYQRO/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/3VKFXRQM6RXARY7J43WUMGYQRO/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":"2026-08-03T17:27:45Z","links":{"resolver":"https://pith.science/pith/3VKFXRQM6RXARY7J43WUMGYQRO","bundle":"https://pith.science/pith/3VKFXRQM6RXARY7J43WUMGYQRO/bundle.json","state":"https://pith.science/pith/3VKFXRQM6RXARY7J43WUMGYQRO/state.json","well_known_bundle":"https://pith.science/.well-known/pith/3VKFXRQM6RXARY7J43WUMGYQRO/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:3VKFXRQM6RXARY7J43WUMGYQRO","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":"8635d78054aeba55bcc86b93793aeb1cb6e41a0c5a13564e20ba2274e366a0c6","cross_cats_sorted":["cond-mat.dis-nn","stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-02T06:55:15Z","title_canon_sha256":"52163603e99629cbdd0bb6cb17146be35167df9f9963c970d5f3e49cf163a903"},"schema_version":"1.0","source":{"id":"2605.01288","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.01288","created_at":"2026-06-24T01:15:03Z"},{"alias_kind":"arxiv_version","alias_value":"2605.01288v3","created_at":"2026-06-24T01:15:03Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.01288","created_at":"2026-06-24T01:15:03Z"},{"alias_kind":"pith_short_12","alias_value":"3VKFXRQM6RXA","created_at":"2026-06-24T01:15:03Z"},{"alias_kind":"pith_short_16","alias_value":"3VKFXRQM6RXARY7J","created_at":"2026-06-24T01:15:03Z"},{"alias_kind":"pith_short_8","alias_value":"3VKFXRQM","created_at":"2026-06-24T01:15:03Z"}],"graph_snapshots":[{"event_id":"sha256:559760e5bc3265b033a9495c06de415d513b608fee65c87974567adb163d349c","target":"graph","created_at":"2026-06-24T01:15:03Z","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 derive an exact identity for the imbalance of Frobenius norms of layer weight matrices that holds for any smooth activation and any differentiable loss [...] giving a critical-depth escape time law τ★ = Θ(ε^{-(r-2)}) governed by the number r of layers at the bottleneck scale rather than the total depth L."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The reduction to a scalar ODE relies on an approximate balance law on the permutation-symmetric submanifold whose accuracy and range of validity are not derived from first principles but stated as holding approximately."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"An exact norm-imbalance identity classifies activations into four classes and reduces deep nonlinear training flow to a scalar ODE that predicts saddle escape time scaling as ε to the power of minus (r-2) for r bottleneck layers."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"An exact identity on Frobenius norm imbalances in layer weights reduces deep nonlinear training to a scalar ODE whose escape time scales as ε to the power of minus (r minus 2), where r is the number of bottleneck layers."}],"snapshot_sha256":"f6bfc38509f467af70de9264d04f9b884fdae7fcd6580e2a244fabe181f13984"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"37a72047ffc37f0b237735c20891583088a0a2d5e11361d6c33f9df2470661db"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-20T18:35:43.254592Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T17:26:57.271652Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.01288/integrity.json","findings":[],"snapshot_sha256":"952980ec58cd8c73efa880119863bdc6b10d930db90d3c8491886652e2a991f4","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"In deep networks with small initialization, training exhibits long plateaus separated by sharp feature-acquisition transitions. Whereas shallow nonlinear networks and deep linear networks are well studied, extending these analyses to deep nonlinear networks remains challenging. We derive an exact identity for the imbalance of Frobenius norms of layer weight matrices that holds for any smooth activation and any differentiable loss and use this to classify activation functions into four universality classes. On the permutation-symmetric submanifold, the identity combines with an approximate bala","authors_text":"Divit Rawal, Michael R. DeWeese","cross_cats":["cond-mat.dis-nn","stat.ML"],"headline":"An exact identity on Frobenius norm imbalances in layer weights reduces deep nonlinear training to a scalar ODE whose escape time scales as ε to the power of minus (r minus 2), where r is the number of bottleneck layers.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-02T06:55:15Z","title":"A Theory of Saddle Escape in Deep Nonlinear Networks"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.01288","kind":"arxiv","version":3},"verdict":{"created_at":"2026-05-11T02:19:07.595331Z","id":"79015d10-f517-4bfd-958a-c773aae68005","model_set":{"reader":"grok-4.3"},"one_line_summary":"An exact norm-imbalance identity classifies activations into four classes and reduces deep nonlinear training flow to a scalar ODE that predicts saddle escape time scaling as ε to the power of minus (r-2) for r bottleneck layers.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"An exact identity on Frobenius norm imbalances in layer weights reduces deep nonlinear training to a scalar ODE whose escape time scales as ε to the power of minus (r minus 2), where r is the number of bottleneck layers.","strongest_claim":"We derive an exact identity for the imbalance of Frobenius norms of layer weight matrices that holds for any smooth activation and any differentiable loss [...] giving a critical-depth escape time law τ★ = Θ(ε^{-(r-2)}) governed by the number r of layers at the bottleneck scale rather than the total depth L.","weakest_assumption":"The reduction to a scalar ODE relies on an approximate balance law on the permutation-symmetric submanifold whose accuracy and range of validity are not derived from first principles but stated as holding approximately."}},"verdict_id":"79015d10-f517-4bfd-958a-c773aae68005"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:c3c64c40e7dce5de1e9a1040c9b79ed12828dbdc3ff98caea5470a9e97661bed","target":"record","created_at":"2026-06-24T01:15:03Z","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":"8635d78054aeba55bcc86b93793aeb1cb6e41a0c5a13564e20ba2274e366a0c6","cross_cats_sorted":["cond-mat.dis-nn","stat.ML"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-02T06:55:15Z","title_canon_sha256":"52163603e99629cbdd0bb6cb17146be35167df9f9963c970d5f3e49cf163a903"},"schema_version":"1.0","source":{"id":"2605.01288","kind":"arxiv","version":3}},"canonical_sha256":"dd545bc60cf46e08e3e9e6ed461b108b829581fa3e9cacc5ccac9f69c012f4bf","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"dd545bc60cf46e08e3e9e6ed461b108b829581fa3e9cacc5ccac9f69c012f4bf","first_computed_at":"2026-06-24T01:15:03.268859Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-24T01:15:03.268859Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"jQa4d9G+psq9z21BNP9BthOQ/bBctVqlvYM7Te89TWjWOLXMaVYb43cjIy9UQ8SnFheUmDrJF9y/DPgFBZ8oAA==","signature_status":"signed_v1","signed_at":"2026-06-24T01:15:03.269368Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.01288","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:c3c64c40e7dce5de1e9a1040c9b79ed12828dbdc3ff98caea5470a9e97661bed","sha256:559760e5bc3265b033a9495c06de415d513b608fee65c87974567adb163d349c"],"state_sha256":"a13aa77b3ceedc71421f69e788d7e921afca58f8113be1af4b3c0686ebb7eeed"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Xlx6ETlNtiJPT1GFPDUBa1By2nuNRmW9uZ5o2xwQxxI553EiDUfBF3gWDPwNNwjlFGrEBYdzAdP8mv/jzX8QBQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-03T17:27:45.308627Z","bundle_sha256":"9c4b1f943b822b76f7aceb50893445837f5e55f96bec0a609b4dab8c6d2e7c3a"}}