{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:632R7P5R64S2MLESS2UIUZAQEG","short_pith_number":"pith:632R7P5R","schema_version":"1.0","canonical_sha256":"f6f51fbfb1f725a62c9296a88a6410218161fa0970a423284229f357ee7dd46c","source":{"kind":"arxiv","id":"2507.11274","version":2},"attestation_state":"computed","paper":{"title":"Fast Last-Iterate Convergence of SGD in the Smooth Interpolation Regime","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Amit Attia, Matan Schliserman, Tomer Koren, Uri Sherman","submitted_at":"2025-07-15T12:52:47Z","abstract_excerpt":"We study population convergence guarantees of stochastic gradient descent (SGD) for smooth convex objectives in the interpolation regime, where the noise at optimum is zero or near zero. The behavior of the last iterate of SGD in this setting -- particularly with large (constant) stepsizes -- has received growing attention in recent years due to implications for the training of over-parameterized models, as well as to analyzing forgetting in continual learning and to understanding the convergence of the randomized Kaczmarz method for solving linear systems. We establish that after $T$ steps of"},"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":"2507.11274","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-07-15T12:52:47Z","cross_cats_sorted":["math.OC","stat.ML"],"title_canon_sha256":"be6ece8e1e13d109f682480f4ffa7a26934f550d571d7584661f7b739f806315","abstract_canon_sha256":"fc0a4e2333f3fcc385ee307536c943251b9f978b215164383a247af90f7c848b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:44:08.957825Z","signature_b64":"x+FuZCUE7hA+fnKkvPYjsqZFjcr/q5dORURYzYaP8EePCJUVWOYxqJrkfPNwpOEoDuAvplawcZrrQu1WKEh+BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f6f51fbfb1f725a62c9296a88a6410218161fa0970a423284229f357ee7dd46c","last_reissued_at":"2026-07-05T11:44:08.957262Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:44:08.957262Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fast Last-Iterate Convergence of SGD in the Smooth Interpolation Regime","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Amit Attia, Matan Schliserman, Tomer Koren, Uri Sherman","submitted_at":"2025-07-15T12:52:47Z","abstract_excerpt":"We study population convergence guarantees of stochastic gradient descent (SGD) for smooth convex objectives in the interpolation regime, where the noise at optimum is zero or near zero. The behavior of the last iterate of SGD in this setting -- particularly with large (constant) stepsizes -- has received growing attention in recent years due to implications for the training of over-parameterized models, as well as to analyzing forgetting in continual learning and to understanding the convergence of the randomized Kaczmarz method for solving linear systems. We establish that after $T$ steps of"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.11274","kind":"arxiv","version":2},"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/2507.11274/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":"2507.11274","created_at":"2026-07-05T11:44:08.957334+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.11274v2","created_at":"2026-07-05T11:44:08.957334+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.11274","created_at":"2026-07-05T11:44:08.957334+00:00"},{"alias_kind":"pith_short_12","alias_value":"632R7P5R64S2","created_at":"2026-07-05T11:44:08.957334+00:00"},{"alias_kind":"pith_short_16","alias_value":"632R7P5R64S2MLES","created_at":"2026-07-05T11:44:08.957334+00:00"},{"alias_kind":"pith_short_8","alias_value":"632R7P5R","created_at":"2026-07-05T11:44:08.957334+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.18609","citing_title":"Perfect Parallelization in Mini-Batch SGD with Classical Momentum Acceleration","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09909","citing_title":"Last-Iterate Convergence of Randomized Kaczmarz and SGD with Greedy Step Size","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13870","citing_title":"Gradient Descent's Last Iterate is Often (slightly) Suboptimal","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/632R7P5R64S2MLESS2UIUZAQEG","json":"https://pith.science/pith/632R7P5R64S2MLESS2UIUZAQEG.json","graph_json":"https://pith.science/api/pith-number/632R7P5R64S2MLESS2UIUZAQEG/graph.json","events_json":"https://pith.science/api/pith-number/632R7P5R64S2MLESS2UIUZAQEG/events.json","paper":"https://pith.science/paper/632R7P5R"},"agent_actions":{"view_html":"https://pith.science/pith/632R7P5R64S2MLESS2UIUZAQEG","download_json":"https://pith.science/pith/632R7P5R64S2MLESS2UIUZAQEG.json","view_paper":"https://pith.science/paper/632R7P5R","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.11274&json=true","fetch_graph":"https://pith.science/api/pith-number/632R7P5R64S2MLESS2UIUZAQEG/graph.json","fetch_events":"https://pith.science/api/pith-number/632R7P5R64S2MLESS2UIUZAQEG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/632R7P5R64S2MLESS2UIUZAQEG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/632R7P5R64S2MLESS2UIUZAQEG/action/storage_attestation","attest_author":"https://pith.science/pith/632R7P5R64S2MLESS2UIUZAQEG/action/author_attestation","sign_citation":"https://pith.science/pith/632R7P5R64S2MLESS2UIUZAQEG/action/citation_signature","submit_replication":"https://pith.science/pith/632R7P5R64S2MLESS2UIUZAQEG/action/replication_record"}},"created_at":"2026-07-05T11:44:08.957334+00:00","updated_at":"2026-07-05T11:44:08.957334+00:00"}