{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:SSMJYBT4KKOSEIISUA45XWASWW","short_pith_number":"pith:SSMJYBT4","schema_version":"1.0","canonical_sha256":"94989c067c529d222112a039dbd812b5acdfed1da0ac0e2b913778dcd2371824","source":{"kind":"arxiv","id":"2503.14477","version":2},"attestation_state":"computed","paper":{"title":"Calibrating Verbal Uncertainty as a Linear Feature to Reduce Hallucinations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Alan Schelten, Anthony Hartshorn, Cheng Zhang, Lei Yu, Nicola Cancedda, Pascale Fung, Yejin Bang, Yeskendir Koishekenov, Ziwei Ji","submitted_at":"2025-03-18T17:51:04Z","abstract_excerpt":"LLMs often adopt an assertive language style also when making false claims. Such ``overconfident hallucinations'' mislead users and erode trust. Achieving the ability to express in language the actual degree of uncertainty around a claim is therefore of great importance. We find that ``verbal uncertainty'' is governed by a single linear feature in the representation space of LLMs, and show that this has only moderate correlation with the actual ``semantic uncertainty'' of the model. We apply this insight and show that (1) the mismatch between semantic and verbal uncertainty is a better predict"},"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":"2503.14477","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-03-18T17:51:04Z","cross_cats_sorted":[],"title_canon_sha256":"5cce34b9dc3e3d6ed98f18767290b464db506f584274e09c37e2dea5b85f5ff8","abstract_canon_sha256":"ad9f7a18f67f90a65fa3dd86ca96e64ef726b1839db5519403f76dd21bd16aea"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:52:39.974386Z","signature_b64":"EV/hNjUGPg7epCSwhHh95AXv5FJK1wAKI37yK6V2bknWD7ZjEY8HWTmp4b4nCmlOI9of9fhf58UqZXrsEo1lBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"94989c067c529d222112a039dbd812b5acdfed1da0ac0e2b913778dcd2371824","last_reissued_at":"2026-07-05T10:52:39.973884Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:52:39.973884Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Calibrating Verbal Uncertainty as a Linear Feature to Reduce Hallucinations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Alan Schelten, Anthony Hartshorn, Cheng Zhang, Lei Yu, Nicola Cancedda, Pascale Fung, Yejin Bang, Yeskendir Koishekenov, Ziwei Ji","submitted_at":"2025-03-18T17:51:04Z","abstract_excerpt":"LLMs often adopt an assertive language style also when making false claims. Such ``overconfident hallucinations'' mislead users and erode trust. Achieving the ability to express in language the actual degree of uncertainty around a claim is therefore of great importance. We find that ``verbal uncertainty'' is governed by a single linear feature in the representation space of LLMs, and show that this has only moderate correlation with the actual ``semantic uncertainty'' of the model. We apply this insight and show that (1) the mismatch between semantic and verbal uncertainty is a better predict"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.14477","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/2503.14477/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":"2503.14477","created_at":"2026-07-05T10:52:39.973942+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.14477v2","created_at":"2026-07-05T10:52:39.973942+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.14477","created_at":"2026-07-05T10:52:39.973942+00:00"},{"alias_kind":"pith_short_12","alias_value":"SSMJYBT4KKOS","created_at":"2026-07-05T10:52:39.973942+00:00"},{"alias_kind":"pith_short_16","alias_value":"SSMJYBT4KKOSEIIS","created_at":"2026-07-05T10:52:39.973942+00:00"},{"alias_kind":"pith_short_8","alias_value":"SSMJYBT4","created_at":"2026-07-05T10:52:39.973942+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06327","citing_title":"Estimating Uncertainty from Reasoning: A Large-Scale Study of Multi- and Crosslingual MCQA Performance in LLMs","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2606.03969","citing_title":"Quantifying Faithful Confidence Expression in Large Reasoning Models","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2606.32032","citing_title":"Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28778","citing_title":"Can LLMs Use Linguistic Uncertainty Markers to Reliably Reflect Intrinsic Confidence?","ref_index":32,"is_internal_anchor":false},{"citing_arxiv_id":"2505.17855","citing_title":"Explaining Sources of Uncertainty in Automated Fact-Checking","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2506.14387","citing_title":"SEAT: Sparse Entity-Aware Tuning for Knowledge Adaptation while Preserving Epistemic Abstention","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2509.09438","citing_title":"GrACE: A Generative Approach to Better Confidence Elicitation and Efficient Test-Time Scaling in Large Language Models","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01428","citing_title":"Hallucinations Undermine Trust; Metacognition is a Way Forward","ref_index":14,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SSMJYBT4KKOSEIISUA45XWASWW","json":"https://pith.science/pith/SSMJYBT4KKOSEIISUA45XWASWW.json","graph_json":"https://pith.science/api/pith-number/SSMJYBT4KKOSEIISUA45XWASWW/graph.json","events_json":"https://pith.science/api/pith-number/SSMJYBT4KKOSEIISUA45XWASWW/events.json","paper":"https://pith.science/paper/SSMJYBT4"},"agent_actions":{"view_html":"https://pith.science/pith/SSMJYBT4KKOSEIISUA45XWASWW","download_json":"https://pith.science/pith/SSMJYBT4KKOSEIISUA45XWASWW.json","view_paper":"https://pith.science/paper/SSMJYBT4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.14477&json=true","fetch_graph":"https://pith.science/api/pith-number/SSMJYBT4KKOSEIISUA45XWASWW/graph.json","fetch_events":"https://pith.science/api/pith-number/SSMJYBT4KKOSEIISUA45XWASWW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SSMJYBT4KKOSEIISUA45XWASWW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SSMJYBT4KKOSEIISUA45XWASWW/action/storage_attestation","attest_author":"https://pith.science/pith/SSMJYBT4KKOSEIISUA45XWASWW/action/author_attestation","sign_citation":"https://pith.science/pith/SSMJYBT4KKOSEIISUA45XWASWW/action/citation_signature","submit_replication":"https://pith.science/pith/SSMJYBT4KKOSEIISUA45XWASWW/action/replication_record"}},"created_at":"2026-07-05T10:52:39.973942+00:00","updated_at":"2026-07-05T10:52:39.973942+00:00"}