{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:7CDTESRNPDGX7T5WFJV647OVYQ","short_pith_number":"pith:7CDTESRN","schema_version":"1.0","canonical_sha256":"f887324a2d78cd7fcfb62a6bee7dd5c421b060c2014fecba100c9425f34ffc4e","source":{"kind":"arxiv","id":"2010.14134","version":3},"attestation_state":"computed","paper":{"title":"Selective Classification Can Magnify Disparities Across Groups","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Ananya Kumar, Erik Jones, Pang Wei Koh, Percy Liang, Shiori Sagawa","submitted_at":"2020-10-27T08:51:30Z","abstract_excerpt":"Selective classification, in which models can abstain on uncertain predictions, is a natural approach to improving accuracy in settings where errors are costly but abstentions are manageable. In this paper, we find that while selective classification can improve average accuracies, it can simultaneously magnify existing accuracy disparities between various groups within a population, especially in the presence of spurious correlations. We observe this behavior consistently across five vision and NLP datasets. Surprisingly, increasing abstentions can even decrease accuracies on some groups. To "},"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":"2010.14134","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-27T08:51:30Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"549905f22ee45dd7a4bed659e4ddcf163f93004c7b4a7ebae65557862d584553","abstract_canon_sha256":"fabffc2ef6eee1393791c5de1bc1cc1359eba0fec3260064f8d45c03f0e73718"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:32:01.526499Z","signature_b64":"s9F6jdb+OEFmIDqOD1MOoxL3nWN/SDpRxnuhd6nNJxODyxcUm3O27jn6cxdx64iMD8mETp8Y2GT3hsa30jopBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f887324a2d78cd7fcfb62a6bee7dd5c421b060c2014fecba100c9425f34ffc4e","last_reissued_at":"2026-07-05T02:32:01.525995Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:32:01.525995Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Selective Classification Can Magnify Disparities Across Groups","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Ananya Kumar, Erik Jones, Pang Wei Koh, Percy Liang, Shiori Sagawa","submitted_at":"2020-10-27T08:51:30Z","abstract_excerpt":"Selective classification, in which models can abstain on uncertain predictions, is a natural approach to improving accuracy in settings where errors are costly but abstentions are manageable. In this paper, we find that while selective classification can improve average accuracies, it can simultaneously magnify existing accuracy disparities between various groups within a population, especially in the presence of spurious correlations. We observe this behavior consistently across five vision and NLP datasets. Surprisingly, increasing abstentions can even decrease accuracies on some groups. To "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.14134","kind":"arxiv","version":3},"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/2010.14134/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":"2010.14134","created_at":"2026-07-05T02:32:01.526058+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.14134v3","created_at":"2026-07-05T02:32:01.526058+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.14134","created_at":"2026-07-05T02:32:01.526058+00:00"},{"alias_kind":"pith_short_12","alias_value":"7CDTESRNPDGX","created_at":"2026-07-05T02:32:01.526058+00:00"},{"alias_kind":"pith_short_16","alias_value":"7CDTESRNPDGX7T5W","created_at":"2026-07-05T02:32:01.526058+00:00"},{"alias_kind":"pith_short_8","alias_value":"7CDTESRN","created_at":"2026-07-05T02:32:01.526058+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.11764","citing_title":"Decomposing the Generalization Gap in PROTAC Activity Prediction: Variance Attribution and the Inter-Laboratory Ceiling","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7CDTESRNPDGX7T5WFJV647OVYQ","json":"https://pith.science/pith/7CDTESRNPDGX7T5WFJV647OVYQ.json","graph_json":"https://pith.science/api/pith-number/7CDTESRNPDGX7T5WFJV647OVYQ/graph.json","events_json":"https://pith.science/api/pith-number/7CDTESRNPDGX7T5WFJV647OVYQ/events.json","paper":"https://pith.science/paper/7CDTESRN"},"agent_actions":{"view_html":"https://pith.science/pith/7CDTESRNPDGX7T5WFJV647OVYQ","download_json":"https://pith.science/pith/7CDTESRNPDGX7T5WFJV647OVYQ.json","view_paper":"https://pith.science/paper/7CDTESRN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.14134&json=true","fetch_graph":"https://pith.science/api/pith-number/7CDTESRNPDGX7T5WFJV647OVYQ/graph.json","fetch_events":"https://pith.science/api/pith-number/7CDTESRNPDGX7T5WFJV647OVYQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7CDTESRNPDGX7T5WFJV647OVYQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7CDTESRNPDGX7T5WFJV647OVYQ/action/storage_attestation","attest_author":"https://pith.science/pith/7CDTESRNPDGX7T5WFJV647OVYQ/action/author_attestation","sign_citation":"https://pith.science/pith/7CDTESRNPDGX7T5WFJV647OVYQ/action/citation_signature","submit_replication":"https://pith.science/pith/7CDTESRNPDGX7T5WFJV647OVYQ/action/replication_record"}},"created_at":"2026-07-05T02:32:01.526058+00:00","updated_at":"2026-07-05T02:32:01.526058+00:00"}