{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:RTYK4HCGRHXC2IJOOBZIVOKAUL","short_pith_number":"pith:RTYK4HCG","schema_version":"1.0","canonical_sha256":"8cf0ae1c4689ee2d212e70728ab940a2d9b06f5641b8fe02452d4ceb3264c083","source":{"kind":"arxiv","id":"2501.19195","version":2},"attestation_state":"computed","paper":{"title":"Rethinking Early Stopping: Refine, Then Calibrate","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"David Holzm\\\"uller, Eug\\`ene Berta, Francis Bach, Michael I. Jordan","submitted_at":"2025-01-31T15:03:54Z","abstract_excerpt":"Machine learning classifiers often produce probabilistic predictions that are critical for accurate and interpretable decision-making in various domains. The quality of these predictions is generally evaluated with proper losses, such as cross-entropy, which decompose into two components: calibration error assesses general under/overconfidence, while refinement error measures the ability to distinguish different classes. In this paper, we present a novel variational formulation of the calibration-refinement decomposition that sheds new light on post-hoc calibration, and enables rapid estimatio"},"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":"2501.19195","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-01-31T15:03:54Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"b64e780cf1377320333bfd83000607c9e7dbc5b04829085a76589c4ae8e9b5d8","abstract_canon_sha256":"8fcaa9ed5a27a77e00357559e75c913b3fb7f38435af4cf575e1d2b1a7dc5e7e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:26:47.305290Z","signature_b64":"XJMGxvXtRLFcytyc86cwLKPI4KondaBYDkT+Ebs/F7AccXkUdwxXjQ0VG4aNZeSL9uIAlYNG6iHz7Je4AVPIBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8cf0ae1c4689ee2d212e70728ab940a2d9b06f5641b8fe02452d4ceb3264c083","last_reissued_at":"2026-07-05T11:26:47.304656Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:26:47.304656Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Rethinking Early Stopping: Refine, Then Calibrate","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"David Holzm\\\"uller, Eug\\`ene Berta, Francis Bach, Michael I. Jordan","submitted_at":"2025-01-31T15:03:54Z","abstract_excerpt":"Machine learning classifiers often produce probabilistic predictions that are critical for accurate and interpretable decision-making in various domains. The quality of these predictions is generally evaluated with proper losses, such as cross-entropy, which decompose into two components: calibration error assesses general under/overconfidence, while refinement error measures the ability to distinguish different classes. In this paper, we present a novel variational formulation of the calibration-refinement decomposition that sheds new light on post-hoc calibration, and enables rapid estimatio"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.19195","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/2501.19195/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":"2501.19195","created_at":"2026-07-05T11:26:47.304730+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.19195v2","created_at":"2026-07-05T11:26:47.304730+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.19195","created_at":"2026-07-05T11:26:47.304730+00:00"},{"alias_kind":"pith_short_12","alias_value":"RTYK4HCGRHXC","created_at":"2026-07-05T11:26:47.304730+00:00"},{"alias_kind":"pith_short_16","alias_value":"RTYK4HCGRHXC2IJO","created_at":"2026-07-05T11:26:47.304730+00:00"},{"alias_kind":"pith_short_8","alias_value":"RTYK4HCG","created_at":"2026-07-05T11:26:47.304730+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.08377","citing_title":"Eigenvalue Calibration for Semantic Embeddings of Large Language Models","ref_index":190,"is_internal_anchor":true},{"citing_arxiv_id":"2605.30188","citing_title":"CalArena: A Large-Scale Post-Hoc Calibration Benchmark","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.20614","citing_title":"Too Sharp, Too Sure: When Calibration Follows Curvature","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RTYK4HCGRHXC2IJOOBZIVOKAUL","json":"https://pith.science/pith/RTYK4HCGRHXC2IJOOBZIVOKAUL.json","graph_json":"https://pith.science/api/pith-number/RTYK4HCGRHXC2IJOOBZIVOKAUL/graph.json","events_json":"https://pith.science/api/pith-number/RTYK4HCGRHXC2IJOOBZIVOKAUL/events.json","paper":"https://pith.science/paper/RTYK4HCG"},"agent_actions":{"view_html":"https://pith.science/pith/RTYK4HCGRHXC2IJOOBZIVOKAUL","download_json":"https://pith.science/pith/RTYK4HCGRHXC2IJOOBZIVOKAUL.json","view_paper":"https://pith.science/paper/RTYK4HCG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.19195&json=true","fetch_graph":"https://pith.science/api/pith-number/RTYK4HCGRHXC2IJOOBZIVOKAUL/graph.json","fetch_events":"https://pith.science/api/pith-number/RTYK4HCGRHXC2IJOOBZIVOKAUL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RTYK4HCGRHXC2IJOOBZIVOKAUL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RTYK4HCGRHXC2IJOOBZIVOKAUL/action/storage_attestation","attest_author":"https://pith.science/pith/RTYK4HCGRHXC2IJOOBZIVOKAUL/action/author_attestation","sign_citation":"https://pith.science/pith/RTYK4HCGRHXC2IJOOBZIVOKAUL/action/citation_signature","submit_replication":"https://pith.science/pith/RTYK4HCGRHXC2IJOOBZIVOKAUL/action/replication_record"}},"created_at":"2026-07-05T11:26:47.304730+00:00","updated_at":"2026-07-05T11:26:47.304730+00:00"}