{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:H5QZ7NT2VC5MWGSGPQADR3YFZB","short_pith_number":"pith:H5QZ7NT2","schema_version":"1.0","canonical_sha256":"3f619fb67aa8bacb1a467c0038ef05c868b302e2d02df59fc290cc9cae10aaae","source":{"kind":"arxiv","id":"2506.21849","version":1},"attestation_state":"computed","paper":{"title":"The Consistency Hypothesis in Uncertainty Quantification for Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Balaji Ganesan, Debarun Bhattacharjya, Junkyu Lee, Katsiaryna Mirylenka, Michael Glass, Nhan H Pham, Quan Xiao, Radu Marinescu","submitted_at":"2025-06-27T01:53:15Z","abstract_excerpt":"Estimating the confidence of large language model (LLM) outputs is essential for real-world applications requiring high user trust. Black-box uncertainty quantification (UQ) methods, relying solely on model API access, have gained popularity due to their practical benefits. In this paper, we examine the implicit assumption behind several UQ methods, which use generation consistency as a proxy for confidence, an idea we formalize as the consistency hypothesis. We introduce three mathematical statements with corresponding statistical tests to capture variations of this hypothesis and metrics 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":"2506.21849","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-27T01:53:15Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"7a781e8863c4c4577b80b9441970a2b4962812b4dde30621cdce41c16a9043e9","abstract_canon_sha256":"94c6e0ad583788f1f14ace9991cddc5263288eac516d6f09c41eb0e64961c25a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:28:02.355805Z","signature_b64":"HF83WaM8IpZ5Pi5yWKYgGWrRZLwdNGnY2Bk6NRZS4uVo6Wq12iDx/wxTHxJ6Y99HFuloImxrQcoedA4wOmJjDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3f619fb67aa8bacb1a467c0038ef05c868b302e2d02df59fc290cc9cae10aaae","last_reissued_at":"2026-07-05T11:28:02.355258Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:28:02.355258Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Consistency Hypothesis in Uncertainty Quantification for Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Balaji Ganesan, Debarun Bhattacharjya, Junkyu Lee, Katsiaryna Mirylenka, Michael Glass, Nhan H Pham, Quan Xiao, Radu Marinescu","submitted_at":"2025-06-27T01:53:15Z","abstract_excerpt":"Estimating the confidence of large language model (LLM) outputs is essential for real-world applications requiring high user trust. Black-box uncertainty quantification (UQ) methods, relying solely on model API access, have gained popularity due to their practical benefits. In this paper, we examine the implicit assumption behind several UQ methods, which use generation consistency as a proxy for confidence, an idea we formalize as the consistency hypothesis. We introduce three mathematical statements with corresponding statistical tests to capture variations of this hypothesis and metrics to "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.21849","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/2506.21849/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":"2506.21849","created_at":"2026-07-05T11:28:02.355341+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.21849v1","created_at":"2026-07-05T11:28:02.355341+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.21849","created_at":"2026-07-05T11:28:02.355341+00:00"},{"alias_kind":"pith_short_12","alias_value":"H5QZ7NT2VC5M","created_at":"2026-07-05T11:28:02.355341+00:00"},{"alias_kind":"pith_short_16","alias_value":"H5QZ7NT2VC5MWGSG","created_at":"2026-07-05T11:28:02.355341+00:00"},{"alias_kind":"pith_short_8","alias_value":"H5QZ7NT2","created_at":"2026-07-05T11:28:02.355341+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/H5QZ7NT2VC5MWGSGPQADR3YFZB","json":"https://pith.science/pith/H5QZ7NT2VC5MWGSGPQADR3YFZB.json","graph_json":"https://pith.science/api/pith-number/H5QZ7NT2VC5MWGSGPQADR3YFZB/graph.json","events_json":"https://pith.science/api/pith-number/H5QZ7NT2VC5MWGSGPQADR3YFZB/events.json","paper":"https://pith.science/paper/H5QZ7NT2"},"agent_actions":{"view_html":"https://pith.science/pith/H5QZ7NT2VC5MWGSGPQADR3YFZB","download_json":"https://pith.science/pith/H5QZ7NT2VC5MWGSGPQADR3YFZB.json","view_paper":"https://pith.science/paper/H5QZ7NT2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.21849&json=true","fetch_graph":"https://pith.science/api/pith-number/H5QZ7NT2VC5MWGSGPQADR3YFZB/graph.json","fetch_events":"https://pith.science/api/pith-number/H5QZ7NT2VC5MWGSGPQADR3YFZB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/H5QZ7NT2VC5MWGSGPQADR3YFZB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/H5QZ7NT2VC5MWGSGPQADR3YFZB/action/storage_attestation","attest_author":"https://pith.science/pith/H5QZ7NT2VC5MWGSGPQADR3YFZB/action/author_attestation","sign_citation":"https://pith.science/pith/H5QZ7NT2VC5MWGSGPQADR3YFZB/action/citation_signature","submit_replication":"https://pith.science/pith/H5QZ7NT2VC5MWGSGPQADR3YFZB/action/replication_record"}},"created_at":"2026-07-05T11:28:02.355341+00:00","updated_at":"2026-07-05T11:28:02.355341+00:00"}