{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ZFD4QSOH5PFM2B3UVRAEMRT4XP","short_pith_number":"pith:ZFD4QSOH","schema_version":"1.0","canonical_sha256":"c947c849c7ebcacd0774ac4046467cbbd907f876c07231d55a8f9dd2fa44fc33","source":{"kind":"arxiv","id":"2607.18086","version":1},"attestation_state":"computed","paper":{"title":"Judge-dependent safety gains and model-specific helpfulness costs of evidence-sufficiency prompting in clinical LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Koyar Afrasyab","submitted_at":"2026-07-20T15:53:02Z","abstract_excerpt":"Background: LLM judges increasingly score whether clinical language models give overconfident answers under incomplete evidence, yet whether a measured \"safety gain\" reflects real behavior change or the judge's calibration is unresolved. Using a structured evidence-sufficiency prompt as a test case, we asked whether it reduces unsafe overconfident answers, how far that effect depends on the scoring judge, and what it costs in helpfulness.\n  Methods: In a retrospective public-data benchmark (Real-POCQi, HealthBench, MedRBench), four models (GPT-5.5, Claude Opus 4.8, Gemini 3.5 Flash, Grok 4.3) "},"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":"2607.18086","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-20T15:53:02Z","cross_cats_sorted":[],"title_canon_sha256":"51f6d374f48f9f1119c52327c442b3639781948e203cf749af40b0540d532f3f","abstract_canon_sha256":"cf91f68f122c9257be426d7a39f1bbd28991d765648fc494126e06ea0e6806af"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-21T02:22:15.678594Z","signature_b64":"YrszHmyERBW0ak6GkDODTQaOjl9uUIOY36LkfetClFhJisw3ZQydxb79x5qSNZevix2rVty/TNVarli52Q1HAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c947c849c7ebcacd0774ac4046467cbbd907f876c07231d55a8f9dd2fa44fc33","last_reissued_at":"2026-07-21T02:22:15.677704Z","signature_status":"signed_v1","first_computed_at":"2026-07-21T02:22:15.677704Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Judge-dependent safety gains and model-specific helpfulness costs of evidence-sufficiency prompting in clinical LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Koyar Afrasyab","submitted_at":"2026-07-20T15:53:02Z","abstract_excerpt":"Background: LLM judges increasingly score whether clinical language models give overconfident answers under incomplete evidence, yet whether a measured \"safety gain\" reflects real behavior change or the judge's calibration is unresolved. Using a structured evidence-sufficiency prompt as a test case, we asked whether it reduces unsafe overconfident answers, how far that effect depends on the scoring judge, and what it costs in helpfulness.\n  Methods: In a retrospective public-data benchmark (Real-POCQi, HealthBench, MedRBench), four models (GPT-5.5, Claude Opus 4.8, Gemini 3.5 Flash, Grok 4.3) "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.18086","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/2607.18086/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":"2607.18086","created_at":"2026-07-21T02:22:15.678166+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.18086v1","created_at":"2026-07-21T02:22:15.678166+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.18086","created_at":"2026-07-21T02:22:15.678166+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZFD4QSOH5PFM","created_at":"2026-07-21T02:22:15.678166+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZFD4QSOH5PFM2B3U","created_at":"2026-07-21T02:22:15.678166+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZFD4QSOH","created_at":"2026-07-21T02:22:15.678166+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/ZFD4QSOH5PFM2B3UVRAEMRT4XP","json":"https://pith.science/pith/ZFD4QSOH5PFM2B3UVRAEMRT4XP.json","graph_json":"https://pith.science/api/pith-number/ZFD4QSOH5PFM2B3UVRAEMRT4XP/graph.json","events_json":"https://pith.science/api/pith-number/ZFD4QSOH5PFM2B3UVRAEMRT4XP/events.json","paper":"https://pith.science/paper/ZFD4QSOH"},"agent_actions":{"view_html":"https://pith.science/pith/ZFD4QSOH5PFM2B3UVRAEMRT4XP","download_json":"https://pith.science/pith/ZFD4QSOH5PFM2B3UVRAEMRT4XP.json","view_paper":"https://pith.science/paper/ZFD4QSOH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.18086&json=true","fetch_graph":"https://pith.science/api/pith-number/ZFD4QSOH5PFM2B3UVRAEMRT4XP/graph.json","fetch_events":"https://pith.science/api/pith-number/ZFD4QSOH5PFM2B3UVRAEMRT4XP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZFD4QSOH5PFM2B3UVRAEMRT4XP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZFD4QSOH5PFM2B3UVRAEMRT4XP/action/storage_attestation","attest_author":"https://pith.science/pith/ZFD4QSOH5PFM2B3UVRAEMRT4XP/action/author_attestation","sign_citation":"https://pith.science/pith/ZFD4QSOH5PFM2B3UVRAEMRT4XP/action/citation_signature","submit_replication":"https://pith.science/pith/ZFD4QSOH5PFM2B3UVRAEMRT4XP/action/replication_record"}},"created_at":"2026-07-21T02:22:15.678166+00:00","updated_at":"2026-07-21T02:22:15.678166+00:00"}