{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:DOQXLNFXMPDLNSRZSOHTEE5CJY","short_pith_number":"pith:DOQXLNFX","schema_version":"1.0","canonical_sha256":"1ba175b4b763c6b6ca39938f3213a24e2f5c1abcd114951115023d19e56d09ac","source":{"kind":"arxiv","id":"2405.01563","version":1},"attestation_state":"computed","paper":{"title":"Mitigating LLM Hallucinations via Conformal Abstention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Adam Fisch, Ali Taylan Cemgil, Andr\\'as Gy\\\"orgy, Arnaud Doucet, Csaba Szepesv\\'ari, David Stutz, Ilja Kuzborskij, Iuliya Beloshapka, Nenad Tomasev, Wei-Hung Weng, Yao-Yuan Yang, Yasin Abbasi Yadkori","submitted_at":"2024-04-04T11:32:03Z","abstract_excerpt":"We develop a principled procedure for determining when a large language model (LLM) should abstain from responding (e.g., by saying \"I don't know\") in a general domain, instead of resorting to possibly \"hallucinating\" a non-sensical or incorrect answer. Building on earlier approaches that use self-consistency as a more reliable measure of model confidence, we propose using the LLM itself to self-evaluate the similarity between each of its sampled responses for a given query. We then further leverage conformal prediction techniques to develop an abstention procedure that benefits from rigorous "},"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":"2405.01563","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-04-04T11:32:03Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"57e2621a505546ab1c58a99396fc6caafc88daee10a70bc4148ff1bcbc0b9f02","abstract_canon_sha256":"e2ca3daa807ab80db7894f6834d2cc0ae626c9c5ae14fe04e6484cf330a37b02"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:14:57.969795Z","signature_b64":"s6L7e1CRtm2rcZxxifmN5CP37NGA0jXy1RtV9vM0t+Bfjuej8aQU0M0xG5ORinE812dc6pfp2IcdmNL6TwlFCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1ba175b4b763c6b6ca39938f3213a24e2f5c1abcd114951115023d19e56d09ac","last_reissued_at":"2026-07-05T08:14:57.969304Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:14:57.969304Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Mitigating LLM Hallucinations via Conformal Abstention","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Adam Fisch, Ali Taylan Cemgil, Andr\\'as Gy\\\"orgy, Arnaud Doucet, Csaba Szepesv\\'ari, David Stutz, Ilja Kuzborskij, Iuliya Beloshapka, Nenad Tomasev, Wei-Hung Weng, Yao-Yuan Yang, Yasin Abbasi Yadkori","submitted_at":"2024-04-04T11:32:03Z","abstract_excerpt":"We develop a principled procedure for determining when a large language model (LLM) should abstain from responding (e.g., by saying \"I don't know\") in a general domain, instead of resorting to possibly \"hallucinating\" a non-sensical or incorrect answer. Building on earlier approaches that use self-consistency as a more reliable measure of model confidence, we propose using the LLM itself to self-evaluate the similarity between each of its sampled responses for a given query. We then further leverage conformal prediction techniques to develop an abstention procedure that benefits from rigorous "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.01563","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/2405.01563/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":"2405.01563","created_at":"2026-07-05T08:14:57.969369+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.01563v1","created_at":"2026-07-05T08:14:57.969369+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.01563","created_at":"2026-07-05T08:14:57.969369+00:00"},{"alias_kind":"pith_short_12","alias_value":"DOQXLNFXMPDL","created_at":"2026-07-05T08:14:57.969369+00:00"},{"alias_kind":"pith_short_16","alias_value":"DOQXLNFXMPDLNSRZ","created_at":"2026-07-05T08:14:57.969369+00:00"},{"alias_kind":"pith_short_8","alias_value":"DOQXLNFX","created_at":"2026-07-05T08:14:57.969369+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":17,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2607.08456","citing_title":"Two Axes of LLM Abstention: Answer Correctness and Question Answerability","ref_index":1,"is_internal_anchor":true},{"citing_arxiv_id":"2607.05985","citing_title":"Auto-DSM Under the Lens: A Black-Box Evaluation Framework for LLM-Based DSM Generation","ref_index":21,"is_internal_anchor":true},{"citing_arxiv_id":"2606.12587","citing_title":"Strategic Decision Support for AI Agents","ref_index":95,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06959","citing_title":"OpenHalDet: A Unified Benchmark for Hallucination Detection across Diverse Generation Scenarios","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24053","citing_title":"Breaking the Chains of Probability: Neutrosophic Logic as a New Framework for Epistemic Uncertainty in Large Language Models","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.30085","citing_title":"Conformal Certification of Reasoning Trace Prefixes","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09875","citing_title":"Integrating Local and Global Entropy for Uncertainty Quantification in LLMs","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23189","citing_title":"Empirical Bayes Conformal Prediction for Vision and Language Models","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2603.22161","citing_title":"Causal Evidence that Language Models use Confidence to Drive Behavior","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18812","citing_title":"PASC: Pipeline-Aware Conformal Prediction with Joint Coverage Guarantees for Multi-Stage NLP and LLM Pipelines","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19220","citing_title":"Position: Uncertainty Quantification in LLMs is Just Unsupervised Clustering","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.15416","citing_title":"Margin-Adaptive Confidence Ranking for Reliable LLM Judgement","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2511.02626","citing_title":"Understanding New-Knowledge-Induced Factual Hallucinations in LLMs: Analysis and Interpretation","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09602","citing_title":"From Scalars to Tensors: Declared Losses Recover Epistemic Distinctions That Neutrosophic Scalars Cannot Express","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14098","citing_title":"Pause and Reflect: Conformal Aggregation for Chain-of-Thought Reasoning","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10351","citing_title":"Foundations of Reliable Inference: Reliability-Efficiency Co-Design","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2604.17843","citing_title":"Learning from AVA: Early Lessons from a Curated and Trustworthy Generative AI for Policy and Development Research","ref_index":117,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DOQXLNFXMPDLNSRZSOHTEE5CJY","json":"https://pith.science/pith/DOQXLNFXMPDLNSRZSOHTEE5CJY.json","graph_json":"https://pith.science/api/pith-number/DOQXLNFXMPDLNSRZSOHTEE5CJY/graph.json","events_json":"https://pith.science/api/pith-number/DOQXLNFXMPDLNSRZSOHTEE5CJY/events.json","paper":"https://pith.science/paper/DOQXLNFX"},"agent_actions":{"view_html":"https://pith.science/pith/DOQXLNFXMPDLNSRZSOHTEE5CJY","download_json":"https://pith.science/pith/DOQXLNFXMPDLNSRZSOHTEE5CJY.json","view_paper":"https://pith.science/paper/DOQXLNFX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.01563&json=true","fetch_graph":"https://pith.science/api/pith-number/DOQXLNFXMPDLNSRZSOHTEE5CJY/graph.json","fetch_events":"https://pith.science/api/pith-number/DOQXLNFXMPDLNSRZSOHTEE5CJY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DOQXLNFXMPDLNSRZSOHTEE5CJY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DOQXLNFXMPDLNSRZSOHTEE5CJY/action/storage_attestation","attest_author":"https://pith.science/pith/DOQXLNFXMPDLNSRZSOHTEE5CJY/action/author_attestation","sign_citation":"https://pith.science/pith/DOQXLNFXMPDLNSRZSOHTEE5CJY/action/citation_signature","submit_replication":"https://pith.science/pith/DOQXLNFXMPDLNSRZSOHTEE5CJY/action/replication_record"}},"created_at":"2026-07-05T08:14:57.969369+00:00","updated_at":"2026-07-05T08:14:57.969369+00:00"}