{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:UNCXMYP2BKE3HQABQIFNCTB2JF","short_pith_number":"pith:UNCXMYP2","schema_version":"1.0","canonical_sha256":"a3457661fa0a89b3c001820ad14c3a497c6c84c6a984563c006d938284203b6c","source":{"kind":"arxiv","id":"2506.17163","version":1},"attestation_state":"computed","paper":{"title":"The MedPerturb Dataset: What Non-Content Perturbations Reveal About Human and Clinical LLM Decision Making","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Abinitha Gourabathina, Marzyeh Ghassemi, Walter Gerych, Yuexing Hao","submitted_at":"2025-06-20T17:09:27Z","abstract_excerpt":"Clinical robustness is critical to the safe deployment of medical Large Language Models (LLMs), but key questions remain about how LLMs and humans may differ in response to the real-world variability typified by clinical settings. To address this, we introduce MedPerturb, a dataset designed to systematically evaluate medical LLMs under controlled perturbations of clinical input. MedPerturb consists of clinical vignettes spanning a range of pathologies, each transformed along three axes: (1) gender modifications (e.g., gender-swapping or gender-removal); (2) style variation (e.g., uncertain phr"},"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":"2506.17163","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-06-20T17:09:27Z","cross_cats_sorted":[],"title_canon_sha256":"5ed9102f39366002da5c58d563ffec62549099bd2c4d441863de9f5055443f8f","abstract_canon_sha256":"2cf17efc5246331ad6e5ae39e79dddcd2c403349c8cd87ad5088f06891811c1e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:24:57.235918Z","signature_b64":"CysZYFZBYMvTlZQviZ4bT82RciFcBvz0haBVG8qkPMDu6Npv78b7Y9pWFSci3YJW5pe+hKV6ywYbYEH7DSNZBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a3457661fa0a89b3c001820ad14c3a497c6c84c6a984563c006d938284203b6c","last_reissued_at":"2026-07-05T11:24:57.235326Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:24:57.235326Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The MedPerturb Dataset: What Non-Content Perturbations Reveal About Human and Clinical LLM Decision Making","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Abinitha Gourabathina, Marzyeh Ghassemi, Walter Gerych, Yuexing Hao","submitted_at":"2025-06-20T17:09:27Z","abstract_excerpt":"Clinical robustness is critical to the safe deployment of medical Large Language Models (LLMs), but key questions remain about how LLMs and humans may differ in response to the real-world variability typified by clinical settings. To address this, we introduce MedPerturb, a dataset designed to systematically evaluate medical LLMs under controlled perturbations of clinical input. MedPerturb consists of clinical vignettes spanning a range of pathologies, each transformed along three axes: (1) gender modifications (e.g., gender-swapping or gender-removal); (2) style variation (e.g., uncertain phr"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.17163","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/2506.17163/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":"2506.17163","created_at":"2026-07-05T11:24:57.235416+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.17163v1","created_at":"2026-07-05T11:24:57.235416+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.17163","created_at":"2026-07-05T11:24:57.235416+00:00"},{"alias_kind":"pith_short_12","alias_value":"UNCXMYP2BKE3","created_at":"2026-07-05T11:24:57.235416+00:00"},{"alias_kind":"pith_short_16","alias_value":"UNCXMYP2BKE3HQAB","created_at":"2026-07-05T11:24:57.235416+00:00"},{"alias_kind":"pith_short_8","alias_value":"UNCXMYP2","created_at":"2026-07-05T11:24:57.235416+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.05678","citing_title":"Chain of Risk: Safety Failures in Large Reasoning Models and Mitigation via Adaptive Multi-Principle Steering","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01048","citing_title":"Compared to What? Baselines and Metrics for Counterfactual Prompting","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UNCXMYP2BKE3HQABQIFNCTB2JF","json":"https://pith.science/pith/UNCXMYP2BKE3HQABQIFNCTB2JF.json","graph_json":"https://pith.science/api/pith-number/UNCXMYP2BKE3HQABQIFNCTB2JF/graph.json","events_json":"https://pith.science/api/pith-number/UNCXMYP2BKE3HQABQIFNCTB2JF/events.json","paper":"https://pith.science/paper/UNCXMYP2"},"agent_actions":{"view_html":"https://pith.science/pith/UNCXMYP2BKE3HQABQIFNCTB2JF","download_json":"https://pith.science/pith/UNCXMYP2BKE3HQABQIFNCTB2JF.json","view_paper":"https://pith.science/paper/UNCXMYP2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.17163&json=true","fetch_graph":"https://pith.science/api/pith-number/UNCXMYP2BKE3HQABQIFNCTB2JF/graph.json","fetch_events":"https://pith.science/api/pith-number/UNCXMYP2BKE3HQABQIFNCTB2JF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UNCXMYP2BKE3HQABQIFNCTB2JF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UNCXMYP2BKE3HQABQIFNCTB2JF/action/storage_attestation","attest_author":"https://pith.science/pith/UNCXMYP2BKE3HQABQIFNCTB2JF/action/author_attestation","sign_citation":"https://pith.science/pith/UNCXMYP2BKE3HQABQIFNCTB2JF/action/citation_signature","submit_replication":"https://pith.science/pith/UNCXMYP2BKE3HQABQIFNCTB2JF/action/replication_record"}},"created_at":"2026-07-05T11:24:57.235416+00:00","updated_at":"2026-07-05T11:24:57.235416+00:00"}