{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:TDHNDURGQNJMJJADV3PO3NDWWY","short_pith_number":"pith:TDHNDURG","schema_version":"1.0","canonical_sha256":"98ced1d2268352c4a403aedeedb476b61050591b79f596af1948275d6fa7c2c7","source":{"kind":"arxiv","id":"2307.02764","version":2},"attestation_state":"computed","paper":{"title":"When Does Confidence-Based Cascade Deferral Suffice?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Aditya Krishna Menon, Ankit Singh Rawat, Harikrishna Narasimhan, Neha Gupta, Sanjiv Kumar, Wittawat Jitkrittum","submitted_at":"2023-07-06T04:13:57Z","abstract_excerpt":"Cascades are a classical strategy to enable inference cost to vary adaptively across samples, wherein a sequence of classifiers are invoked in turn. A deferral rule determines whether to invoke the next classifier in the sequence, or to terminate prediction. One simple deferral rule employs the confidence of the current classifier, e.g., based on the maximum predicted softmax probability. Despite being oblivious to the structure of the cascade -- e.g., not modelling the errors of downstream models -- such confidence-based deferral often works remarkably well in practice. In this paper, we seek"},"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":"2307.02764","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-07-06T04:13:57Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"87a447c53c64ad7f8a749819d328061395dcfe18632b9668d193df2acdbf2926","abstract_canon_sha256":"af2b44b2199b1dabb4935d42aeee8459346d5eba2c3d36d868ef21fc4f3e102b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:36:22.289992Z","signature_b64":"RJd3JNhdbJNATMZexAOOM547hh4PpbDP8HZNFGfpMJ2S6w22oucGZ4xgROPq9leRJe5xZjzERr1MSW+4KGltDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"98ced1d2268352c4a403aedeedb476b61050591b79f596af1948275d6fa7c2c7","last_reissued_at":"2026-07-05T07:36:22.289513Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:36:22.289513Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When Does Confidence-Based Cascade Deferral Suffice?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Aditya Krishna Menon, Ankit Singh Rawat, Harikrishna Narasimhan, Neha Gupta, Sanjiv Kumar, Wittawat Jitkrittum","submitted_at":"2023-07-06T04:13:57Z","abstract_excerpt":"Cascades are a classical strategy to enable inference cost to vary adaptively across samples, wherein a sequence of classifiers are invoked in turn. A deferral rule determines whether to invoke the next classifier in the sequence, or to terminate prediction. One simple deferral rule employs the confidence of the current classifier, e.g., based on the maximum predicted softmax probability. Despite being oblivious to the structure of the cascade -- e.g., not modelling the errors of downstream models -- such confidence-based deferral often works remarkably well in practice. In this paper, we seek"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2307.02764","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/2307.02764/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":"2307.02764","created_at":"2026-07-05T07:36:22.289570+00:00"},{"alias_kind":"arxiv_version","alias_value":"2307.02764v2","created_at":"2026-07-05T07:36:22.289570+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2307.02764","created_at":"2026-07-05T07:36:22.289570+00:00"},{"alias_kind":"pith_short_12","alias_value":"TDHNDURGQNJM","created_at":"2026-07-05T07:36:22.289570+00:00"},{"alias_kind":"pith_short_16","alias_value":"TDHNDURGQNJMJJAD","created_at":"2026-07-05T07:36:22.289570+00:00"},{"alias_kind":"pith_short_8","alias_value":"TDHNDURG","created_at":"2026-07-05T07:36:22.289570+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27288","citing_title":"When Does Combining Language Models Help? A Co-Failure Ceiling on Routing, Voting, and Mixture-of-Agents Across 67 Frontier Models","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.12303","citing_title":"From 2D Grids to 1D Tokens: Reforming Shared Representations for Multimodal Image Fusion","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31331","citing_title":"Expected Gain-based Escalation in Vertical Federated Learning","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2410.15761","citing_title":"Optimal Query Allocation in Extractive QA with LLMs: A Learning-to-Defer Framework with Theoretical Guarantees","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06339","citing_title":"A Regime Theory of Controller Class Selection for LLM Action Decisions","ref_index":16,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TDHNDURGQNJMJJADV3PO3NDWWY","json":"https://pith.science/pith/TDHNDURGQNJMJJADV3PO3NDWWY.json","graph_json":"https://pith.science/api/pith-number/TDHNDURGQNJMJJADV3PO3NDWWY/graph.json","events_json":"https://pith.science/api/pith-number/TDHNDURGQNJMJJADV3PO3NDWWY/events.json","paper":"https://pith.science/paper/TDHNDURG"},"agent_actions":{"view_html":"https://pith.science/pith/TDHNDURGQNJMJJADV3PO3NDWWY","download_json":"https://pith.science/pith/TDHNDURGQNJMJJADV3PO3NDWWY.json","view_paper":"https://pith.science/paper/TDHNDURG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2307.02764&json=true","fetch_graph":"https://pith.science/api/pith-number/TDHNDURGQNJMJJADV3PO3NDWWY/graph.json","fetch_events":"https://pith.science/api/pith-number/TDHNDURGQNJMJJADV3PO3NDWWY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TDHNDURGQNJMJJADV3PO3NDWWY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TDHNDURGQNJMJJADV3PO3NDWWY/action/storage_attestation","attest_author":"https://pith.science/pith/TDHNDURGQNJMJJADV3PO3NDWWY/action/author_attestation","sign_citation":"https://pith.science/pith/TDHNDURGQNJMJJADV3PO3NDWWY/action/citation_signature","submit_replication":"https://pith.science/pith/TDHNDURGQNJMJJADV3PO3NDWWY/action/replication_record"}},"created_at":"2026-07-05T07:36:22.289570+00:00","updated_at":"2026-07-05T07:36:22.289570+00:00"}