{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:TDQ2DZIPB7AZ7NCAGMFJKJDJZM","short_pith_number":"pith:TDQ2DZIP","schema_version":"1.0","canonical_sha256":"98e1a1e50f0fc19fb440330a952469cb11619cb01477c0ea703ee9d64b74d25c","source":{"kind":"arxiv","id":"2505.22359","version":1},"attestation_state":"computed","paper":{"title":"Multiclass Loss Geometry Matters for Generalization of Gradient Descent in Separable Classification","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Matan Schliserman, Tomer Koren","submitted_at":"2025-05-28T13:39:14Z","abstract_excerpt":"We study the generalization performance of unregularized gradient methods for separable linear classification. While previous work mostly deal with the binary case, we focus on the multiclass setting with $k$ classes and establish novel population risk bounds for Gradient Descent for loss functions that decay to zero. In this setting, we show risk bounds that reveal that convergence rates are crucially influenced by the geometry of the loss template, as formalized by Wang and Scott (2024), rather than of the loss function itself. Particularly, we establish risk upper bounds that holds for any "},"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":"2505.22359","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-05-28T13:39:14Z","cross_cats_sorted":[],"title_canon_sha256":"4706c5dee54864c76bb62afedcce3ae7bcbd236c2320d2384f8edf59ab9361ef","abstract_canon_sha256":"7389cb8422c0468b320ecc95223be7f8c02e717c985bbb0cc572791614a29847"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:11:13.864327Z","signature_b64":"cFVCOhSdOX3azWU1Ro2BvOewTQ6o3mVvW+3C2gMBQKOC3YCqjwT3M9Dko7Mx9t7n6bg1Hren76Na3WS+P3VxAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"98e1a1e50f0fc19fb440330a952469cb11619cb01477c0ea703ee9d64b74d25c","last_reissued_at":"2026-07-05T11:11:13.863830Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:11:13.863830Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multiclass Loss Geometry Matters for Generalization of Gradient Descent in Separable Classification","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Matan Schliserman, Tomer Koren","submitted_at":"2025-05-28T13:39:14Z","abstract_excerpt":"We study the generalization performance of unregularized gradient methods for separable linear classification. While previous work mostly deal with the binary case, we focus on the multiclass setting with $k$ classes and establish novel population risk bounds for Gradient Descent for loss functions that decay to zero. In this setting, we show risk bounds that reveal that convergence rates are crucially influenced by the geometry of the loss template, as formalized by Wang and Scott (2024), rather than of the loss function itself. Particularly, we establish risk upper bounds that holds for any "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.22359","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/2505.22359/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":"2505.22359","created_at":"2026-07-05T11:11:13.863888+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.22359v1","created_at":"2026-07-05T11:11:13.863888+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.22359","created_at":"2026-07-05T11:11:13.863888+00:00"},{"alias_kind":"pith_short_12","alias_value":"TDQ2DZIPB7AZ","created_at":"2026-07-05T11:11:13.863888+00:00"},{"alias_kind":"pith_short_16","alias_value":"TDQ2DZIPB7AZ7NCA","created_at":"2026-07-05T11:11:13.863888+00:00"},{"alias_kind":"pith_short_8","alias_value":"TDQ2DZIP","created_at":"2026-07-05T11:11:13.863888+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/TDQ2DZIPB7AZ7NCAGMFJKJDJZM","json":"https://pith.science/pith/TDQ2DZIPB7AZ7NCAGMFJKJDJZM.json","graph_json":"https://pith.science/api/pith-number/TDQ2DZIPB7AZ7NCAGMFJKJDJZM/graph.json","events_json":"https://pith.science/api/pith-number/TDQ2DZIPB7AZ7NCAGMFJKJDJZM/events.json","paper":"https://pith.science/paper/TDQ2DZIP"},"agent_actions":{"view_html":"https://pith.science/pith/TDQ2DZIPB7AZ7NCAGMFJKJDJZM","download_json":"https://pith.science/pith/TDQ2DZIPB7AZ7NCAGMFJKJDJZM.json","view_paper":"https://pith.science/paper/TDQ2DZIP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.22359&json=true","fetch_graph":"https://pith.science/api/pith-number/TDQ2DZIPB7AZ7NCAGMFJKJDJZM/graph.json","fetch_events":"https://pith.science/api/pith-number/TDQ2DZIPB7AZ7NCAGMFJKJDJZM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TDQ2DZIPB7AZ7NCAGMFJKJDJZM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TDQ2DZIPB7AZ7NCAGMFJKJDJZM/action/storage_attestation","attest_author":"https://pith.science/pith/TDQ2DZIPB7AZ7NCAGMFJKJDJZM/action/author_attestation","sign_citation":"https://pith.science/pith/TDQ2DZIPB7AZ7NCAGMFJKJDJZM/action/citation_signature","submit_replication":"https://pith.science/pith/TDQ2DZIPB7AZ7NCAGMFJKJDJZM/action/replication_record"}},"created_at":"2026-07-05T11:11:13.863888+00:00","updated_at":"2026-07-05T11:11:13.863888+00:00"}