{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:VQMPFD7EHMUMO6FVYUPLBU6POS","merge_version":"pith-open-graph-merge-v1","event_count":4,"valid_event_count":4,"invalid_event_count":0,"equivocation_count":1,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"133a9adea01e80b03ff81c4967815dba5865719a1437782d91a280f73ba236a3","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-08T15:55:30Z","title_canon_sha256":"53ac9af92f70e8a5751c4d6922585835fb90690dc00399e153d13ad726421550"},"schema_version":"1.0","source":{"id":"2605.07919","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.07919","created_at":"2026-05-25T02:02:16Z"},{"alias_kind":"arxiv_version","alias_value":"2605.07919v2","created_at":"2026-05-25T02:02:16Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.07919","created_at":"2026-05-25T02:02:16Z"},{"alias_kind":"pith_short_12","alias_value":"VQMPFD7EHMUM","created_at":"2026-05-25T02:02:16Z"},{"alias_kind":"pith_short_16","alias_value":"VQMPFD7EHMUMO6FV","created_at":"2026-05-25T02:02:16Z"},{"alias_kind":"pith_short_8","alias_value":"VQMPFD7E","created_at":"2026-05-25T02:02:16Z"}],"graph_snapshots":[{"event_id":"sha256:c3d4bac97121078b6e87557fb6ab519e093feb65cc4ce1c771e857c65c0ac20c","target":"graph","created_at":"2026-05-25T02:02:16Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"The independent radiologist scores MCS 83.3 at silent-failure rate 5.8%, leaving a 14.1-point composite headroom above the strongest audited model (Claude Opus 4.7 at 69.2)."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the four chosen perturbation types and the 300 clinician-authored cases sufficiently represent the range of broken visual evidence that occurs in real clinical practice."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"MedVIGIL introduces a clinician-supervised benchmark showing medical VLMs frequently give fluent answers on broken visual evidence, with top models 14 points below human radiologists on the composite score."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Medical vision-language models fail to refuse answers when visual evidence is broken, trailing radiologists by 14 points on a new composite score."}],"snapshot_sha256":"5abf56945b417f28f53c287fd507a3ccfd3d0c308bedf4408748d780b5d4b2a2"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":false,"detectors_run":[{"findings_count":0,"name":"claim_evidence","ran_at":"2026-05-20T10:02:13.555939Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-20T04:46:22.489233Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T15:31:18.310769Z","status":"completed","version":"1.0.0"},{"findings_count":2,"name":"doi_compliance","ran_at":"2026-05-19T11:24:20.741484Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.07919/integrity.json","findings":[{"audited_at":"2026-05-19T11:24:20.741484Z","detected_arxiv_id":null,"detected_doi":"10.1109/TBME.2021.3117407.URLhttps://pmc.ncbi.nlm.nih.gov/articles/PMC9011180/.Dan","detector":"doi_compliance","finding_type":"recoverable_identifier","note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.1109/TBME.2021.3117407.URLhttps://pmc.ncbi.nlm.nih.gov/articles/PMC9011180/.Dan) was visible in the surrounding text but could not be confirmed against doi.org as printed.","ref_index":10,"severity":"advisory","verdict_class":"incontrovertible"},{"audited_at":"2026-05-19T11:24:20.741484Z","detected_arxiv_id":null,"detected_doi":"10.18653/v1/2023.emnlp-main.20.URLhttps://aclanthology.org/2023.emnlp-main.20/.Bo","detector":"doi_compliance","finding_type":"recoverable_identifier","note":"DOI in the printed bibliography is fragmented by whitespace or line breaks. A longer candidate (10.18653/v1/2023.emnlp-main.20.URLhttps://aclanthology.org/2023.emnlp-main.20/.Bo) was visible in the surrounding text but could not be confirmed against doi.org as printed.","ref_index":17,"severity":"advisory","verdict_class":"incontrovertible"}],"snapshot_sha256":"4abdae5894b4b9e13c60c8601c856f602b2de3eb7fdd7894a72fce302b49f1cc","summary":{"advisory":2,"by_detector":{"doi_compliance":{"advisory":2,"critical":0,"informational":0,"total":2}},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Medical vision--language models (VLMs) are usually evaluated on intact image--question pairs, but trustworthy clinical use requires a stronger property: a model must recognise when the evidential basis for an answer has failed. We study this through silent failures under perturbed evidence, where a vision-required medical question is paired with a false premise, wording perturbation, knowledge-only rewrite, or ROI-corrupted image, yet the model returns a fluent non-refusal answer. We introduce medvigil, a 300-case evaluation suite drawn from four public medical VQA sources, supervised end to e","authors_text":"Hanqi Jiang, Haozhen Gong, Hui Ren, Hyeokjae Kwon, Jinglei Lv, Junhao Chen, Lifeng Chen, Lin Zhao, Mingyu Kang, Quanzheng Li, Ruiyu Yan, Tianming Liu, Weihang You, Xiang Li, Yi Pan","cross_cats":[],"headline":"Medical vision-language models fail to refuse answers when visual evidence is broken, trailing radiologists by 14 points on a new composite score.","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-08T15:55:30Z","title":"MedVIGIL: Evaluating Trustworthy Medical VLMs Under Broken Visual Evidence"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.07919","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-11T03:40:45.192215Z","id":"40477877-05fe-4735-b79f-76bbcd099224","model_set":{"reader":"grok-4.3"},"one_line_summary":"MedVIGIL introduces a clinician-supervised benchmark showing medical VLMs frequently give fluent answers on broken visual evidence, with top models 14 points below human radiologists on the composite score.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Medical vision-language models fail to refuse answers when visual evidence is broken, trailing radiologists by 14 points on a new composite score.","strongest_claim":"The independent radiologist scores MCS 83.3 at silent-failure rate 5.8%, leaving a 14.1-point composite headroom above the strongest audited model (Claude Opus 4.7 at 69.2).","weakest_assumption":"That the four chosen perturbation types and the 300 clinician-authored cases sufficiently represent the range of broken visual evidence that occurs in real clinical practice."}},"verdict_id":"40477877-05fe-4735-b79f-76bbcd099224"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:d987f550ae76bdb920390ab8c740391d590ce01f56262d4c9ee595600804d93c","target":"record","created_at":"2026-05-25T02:02:16Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"133a9adea01e80b03ff81c4967815dba5865719a1437782d91a280f73ba236a3","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-08T15:55:30Z","title_canon_sha256":"53ac9af92f70e8a5751c4d6922585835fb90690dc00399e153d13ad726421550"},"schema_version":"1.0","source":{"id":"2605.07919","kind":"arxiv","version":2}},"canonical_sha256":"ac18f28fe43b28c778b5c51eb0d3cf74b15b8cf4701039e8417b656176d45bf0","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"ac18f28fe43b28c778b5c51eb0d3cf74b15b8cf4701039e8417b656176d45bf0","first_computed_at":"2026-05-25T02:02:16.508920Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-25T02:02:16.508920Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"vAlHsULuIJSPBomTOcKUcnXkUxCkUfgsQcWyKLCbEvTHKByTIlHvd0vHkhGqKT0K8tc9Usamc06CS1fzRGrVDA==","signature_status":"signed_v1","signed_at":"2026-05-25T02:02:16.509641Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.07919","source_kind":"arxiv","source_version":2}}},"equivocations":[{"signer_id":"pith.science","event_type":"integrity_finding","target":"integrity","event_ids":["sha256:9b8ad3dde762e54f4087e169628a6081c445719b4d17664fcb9de725e56fe71c","sha256:c7932fd0ad5b6493465116afc44e150f7ff7f89cde51d18a76123ed4553b00f1"]}],"invalid_events":[],"applied_event_ids":["sha256:d987f550ae76bdb920390ab8c740391d590ce01f56262d4c9ee595600804d93c","sha256:c3d4bac97121078b6e87557fb6ab519e093feb65cc4ce1c771e857c65c0ac20c"],"state_sha256":"30b14fb1cf434e7e2207fc8513d795fab3b2aeb68e22839216535fceba165f33"}