{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:5VDCP2XJAU7TJXCIVISWQXCZ4L","short_pith_number":"pith:5VDCP2XJ","schema_version":"1.0","canonical_sha256":"ed4627eae9053f34dc48aa25685c59e2e502586bf6ef7d2ceb91964e2fe20020","source":{"kind":"arxiv","id":"2607.25018","version":1},"attestation_state":"computed","paper":{"title":"Conformal Cascade: Distribution-Free Accuracy Guarantees for Multi-Tier LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Shibo Li, Shikan Fang, Yifan Dou","submitted_at":"2026-07-27T19:19:58Z","abstract_excerpt":"Large language model (LLM) cascades reduce inference cost by routing easy queries to a small model and deferring hard queries to a larger one. Production cascades govern this deferral through a confidence threshold, but LLM confidence scores are miscalibrated, the threshold must be tuned per model pair and per domain, and no setting yields a formal bound on cascade accuracy. We introduce \\textbf{Conformal Cascade} (CC), a multi-tier inference framework that uses conformal prediction set size as the deferral rule: accept when the calibrated set collapses to a single answer, defer otherwise. The"},"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":"2607.25018","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-27T19:19:58Z","cross_cats_sorted":[],"title_canon_sha256":"b4865a4fdc5aca71e61ef2383d873646b1057029e3348bb0d5a148884bbe1738","abstract_canon_sha256":"3d30a8e68ed87c9a6813e69e80647f40d99e24d69e24c36acc808020ce45b291"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-29T00:24:57.928776Z","signature_b64":"G3l75AOb7DY8EyP0J2gtOoZTdd0wUsk0WXUCeiEYU5MB81rRy6YEwNxCWJXHlWtnC6NbqBFMQU5Cqt+48SacAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ed4627eae9053f34dc48aa25685c59e2e502586bf6ef7d2ceb91964e2fe20020","last_reissued_at":"2026-07-29T00:24:57.927917Z","signature_status":"signed_v1","first_computed_at":"2026-07-29T00:24:57.927917Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Conformal Cascade: Distribution-Free Accuracy Guarantees for Multi-Tier LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Shibo Li, Shikan Fang, Yifan Dou","submitted_at":"2026-07-27T19:19:58Z","abstract_excerpt":"Large language model (LLM) cascades reduce inference cost by routing easy queries to a small model and deferring hard queries to a larger one. Production cascades govern this deferral through a confidence threshold, but LLM confidence scores are miscalibrated, the threshold must be tuned per model pair and per domain, and no setting yields a formal bound on cascade accuracy. We introduce \\textbf{Conformal Cascade} (CC), a multi-tier inference framework that uses conformal prediction set size as the deferral rule: accept when the calibrated set collapses to a single answer, defer otherwise. The"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.25018","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/2607.25018/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":"2607.25018","created_at":"2026-07-29T00:24:57.928360+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.25018v1","created_at":"2026-07-29T00:24:57.928360+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.25018","created_at":"2026-07-29T00:24:57.928360+00:00"},{"alias_kind":"pith_short_12","alias_value":"5VDCP2XJAU7T","created_at":"2026-07-29T00:24:57.928360+00:00"},{"alias_kind":"pith_short_16","alias_value":"5VDCP2XJAU7TJXCI","created_at":"2026-07-29T00:24:57.928360+00:00"},{"alias_kind":"pith_short_8","alias_value":"5VDCP2XJ","created_at":"2026-07-29T00:24:57.928360+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/5VDCP2XJAU7TJXCIVISWQXCZ4L","json":"https://pith.science/pith/5VDCP2XJAU7TJXCIVISWQXCZ4L.json","graph_json":"https://pith.science/api/pith-number/5VDCP2XJAU7TJXCIVISWQXCZ4L/graph.json","events_json":"https://pith.science/api/pith-number/5VDCP2XJAU7TJXCIVISWQXCZ4L/events.json","paper":"https://pith.science/paper/5VDCP2XJ"},"agent_actions":{"view_html":"https://pith.science/pith/5VDCP2XJAU7TJXCIVISWQXCZ4L","download_json":"https://pith.science/pith/5VDCP2XJAU7TJXCIVISWQXCZ4L.json","view_paper":"https://pith.science/paper/5VDCP2XJ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.25018&json=true","fetch_graph":"https://pith.science/api/pith-number/5VDCP2XJAU7TJXCIVISWQXCZ4L/graph.json","fetch_events":"https://pith.science/api/pith-number/5VDCP2XJAU7TJXCIVISWQXCZ4L/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5VDCP2XJAU7TJXCIVISWQXCZ4L/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5VDCP2XJAU7TJXCIVISWQXCZ4L/action/storage_attestation","attest_author":"https://pith.science/pith/5VDCP2XJAU7TJXCIVISWQXCZ4L/action/author_attestation","sign_citation":"https://pith.science/pith/5VDCP2XJAU7TJXCIVISWQXCZ4L/action/citation_signature","submit_replication":"https://pith.science/pith/5VDCP2XJAU7TJXCIVISWQXCZ4L/action/replication_record"}},"created_at":"2026-07-29T00:24:57.928360+00:00","updated_at":"2026-07-29T00:24:57.928360+00:00"}