{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:YATUSZKTMSRZN4DAKMSIGCWJKB","short_pith_number":"pith:YATUSZKT","schema_version":"1.0","canonical_sha256":"c02749655364a396f0605324830ac9507af8b5afa9424da3d57392792090a7f6","source":{"kind":"arxiv","id":"2606.18080","version":1},"attestation_state":"computed","paper":{"title":"Edge Flow: A Tractable and Predictive Continuous-Time Model for Gradient Descent at the Edge of Stability","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Pierre Marion","submitted_at":"2026-06-16T15:45:47Z","abstract_excerpt":"Gradient descent in deep learning may operate at the edge of stability (EoS), a regime in which the largest eigenvalue of the loss Hessian hovers near the stability threshold $2/\\eta$, where $\\eta$ is the learning rate. Classical analysis tools such as gradient flow and the descent lemma do not apply here, motivating the search for a continuous-time model valid at EoS. We propose Edge Flow, a system of three coupled ordinary differential equations that provides a tractable, faithful, and predictive model of gradient descent dynamics at EoS. Edge Flow decomposes the dynamics into a center, an o"},"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":"2606.18080","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-16T15:45:47Z","cross_cats_sorted":[],"title_canon_sha256":"4637b2cf86a781af97ff111a3bf512ec396e190d34380f368359eeac05f34af2","abstract_canon_sha256":"4186dcc38651c0d3aa325957ea974d2690772ca79adc55c3849a29d0a8b9a733"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:10:47.994990Z","signature_b64":"Vg4IXmKiwkNhAaQXMs1/TyN9EMzkNb6iftIfnOEqGPRbpCDK5Ez+chJpJQtT/0+cUsSB3xcsu2xseArfMuF/AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c02749655364a396f0605324830ac9507af8b5afa9424da3d57392792090a7f6","last_reissued_at":"2026-06-19T16:10:47.994648Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:10:47.994648Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Edge Flow: A Tractable and Predictive Continuous-Time Model for Gradient Descent at the Edge of Stability","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Pierre Marion","submitted_at":"2026-06-16T15:45:47Z","abstract_excerpt":"Gradient descent in deep learning may operate at the edge of stability (EoS), a regime in which the largest eigenvalue of the loss Hessian hovers near the stability threshold $2/\\eta$, where $\\eta$ is the learning rate. Classical analysis tools such as gradient flow and the descent lemma do not apply here, motivating the search for a continuous-time model valid at EoS. We propose Edge Flow, a system of three coupled ordinary differential equations that provides a tractable, faithful, and predictive model of gradient descent dynamics at EoS. Edge Flow decomposes the dynamics into a center, an o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.18080","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/2606.18080/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":"2606.18080","created_at":"2026-06-19T16:10:47.994709+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.18080v1","created_at":"2026-06-19T16:10:47.994709+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.18080","created_at":"2026-06-19T16:10:47.994709+00:00"},{"alias_kind":"pith_short_12","alias_value":"YATUSZKTMSRZ","created_at":"2026-06-19T16:10:47.994709+00:00"},{"alias_kind":"pith_short_16","alias_value":"YATUSZKTMSRZN4DA","created_at":"2026-06-19T16:10:47.994709+00:00"},{"alias_kind":"pith_short_8","alias_value":"YATUSZKT","created_at":"2026-06-19T16:10:47.994709+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/YATUSZKTMSRZN4DAKMSIGCWJKB","json":"https://pith.science/pith/YATUSZKTMSRZN4DAKMSIGCWJKB.json","graph_json":"https://pith.science/api/pith-number/YATUSZKTMSRZN4DAKMSIGCWJKB/graph.json","events_json":"https://pith.science/api/pith-number/YATUSZKTMSRZN4DAKMSIGCWJKB/events.json","paper":"https://pith.science/paper/YATUSZKT"},"agent_actions":{"view_html":"https://pith.science/pith/YATUSZKTMSRZN4DAKMSIGCWJKB","download_json":"https://pith.science/pith/YATUSZKTMSRZN4DAKMSIGCWJKB.json","view_paper":"https://pith.science/paper/YATUSZKT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.18080&json=true","fetch_graph":"https://pith.science/api/pith-number/YATUSZKTMSRZN4DAKMSIGCWJKB/graph.json","fetch_events":"https://pith.science/api/pith-number/YATUSZKTMSRZN4DAKMSIGCWJKB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YATUSZKTMSRZN4DAKMSIGCWJKB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YATUSZKTMSRZN4DAKMSIGCWJKB/action/storage_attestation","attest_author":"https://pith.science/pith/YATUSZKTMSRZN4DAKMSIGCWJKB/action/author_attestation","sign_citation":"https://pith.science/pith/YATUSZKTMSRZN4DAKMSIGCWJKB/action/citation_signature","submit_replication":"https://pith.science/pith/YATUSZKTMSRZN4DAKMSIGCWJKB/action/replication_record"}},"created_at":"2026-06-19T16:10:47.994709+00:00","updated_at":"2026-06-19T16:10:47.994709+00:00"}