{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5TUPF2YVEO6A3VAFZKAUUFVQFE","short_pith_number":"pith:5TUPF2YV","schema_version":"1.0","canonical_sha256":"ece8f2eb1523bc0dd405ca814a16b02937cf79c35468c12ca944a0f0f1cfd104","source":{"kind":"arxiv","id":"2507.11200","version":2},"attestation_state":"computed","paper":{"title":"How Far Have Medical Vision-Language Models Come? A Comprehensive Benchmarking Study","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Che Liu, Daniel Rueckert, Jiazhen Pan, Rossella Arcucci, Weixiang Shen, Wenjia Bai","submitted_at":"2025-07-15T11:12:39Z","abstract_excerpt":"Vision-Language Models (VLMs) trained on web-scale corpora excel at natural image tasks and are increasingly repurposed for healthcare; however, their competence in medical tasks remains underexplored. We present a comprehensive evaluation of open-source general-purpose and medically specialised VLMs, ranging from 3B to 72B parameters, across eight benchmarks: MedXpert, OmniMedVQA, PMC-VQA, PathVQA, MMMU, SLAKE, and VQA-RAD. To observe model performance across different aspects, we first separate it into understanding and reasoning components. Three salient findings emerge. First, large genera"},"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":"2507.11200","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-07-15T11:12:39Z","cross_cats_sorted":[],"title_canon_sha256":"6a5f11d0d512081ed3962d49e70a047f48796ea3bdbe5ebd5b3decfc5ab25dfe","abstract_canon_sha256":"787ae56fb4f0d1426da2810caac87412fc93b75c7678c1a0808c0d053e4602fb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:39:14.010807Z","signature_b64":"dl4zAZZ95y81ZobgmNJyjH14B7Q9AIrdE8OB5gPodVCWfLywbc4DtxDEbxpNboAn0alt4gZRG6M8+GVc8At2Dg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ece8f2eb1523bc0dd405ca814a16b02937cf79c35468c12ca944a0f0f1cfd104","last_reissued_at":"2026-07-05T11:39:14.010345Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:39:14.010345Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Far Have Medical Vision-Language Models Come? A Comprehensive Benchmarking Study","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Che Liu, Daniel Rueckert, Jiazhen Pan, Rossella Arcucci, Weixiang Shen, Wenjia Bai","submitted_at":"2025-07-15T11:12:39Z","abstract_excerpt":"Vision-Language Models (VLMs) trained on web-scale corpora excel at natural image tasks and are increasingly repurposed for healthcare; however, their competence in medical tasks remains underexplored. We present a comprehensive evaluation of open-source general-purpose and medically specialised VLMs, ranging from 3B to 72B parameters, across eight benchmarks: MedXpert, OmniMedVQA, PMC-VQA, PathVQA, MMMU, SLAKE, and VQA-RAD. To observe model performance across different aspects, we first separate it into understanding and reasoning components. Three salient findings emerge. First, large genera"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.11200","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/2507.11200/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":"2507.11200","created_at":"2026-07-05T11:39:14.010400+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.11200v2","created_at":"2026-07-05T11:39:14.010400+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.11200","created_at":"2026-07-05T11:39:14.010400+00:00"},{"alias_kind":"pith_short_12","alias_value":"5TUPF2YVEO6A","created_at":"2026-07-05T11:39:14.010400+00:00"},{"alias_kind":"pith_short_16","alias_value":"5TUPF2YVEO6A3VAF","created_at":"2026-07-05T11:39:14.010400+00:00"},{"alias_kind":"pith_short_8","alias_value":"5TUPF2YV","created_at":"2026-07-05T11:39:14.010400+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20723","citing_title":"Evaluation of Medical Vision Language Models HuluMed and MedGemma, and general purpose chatbots Gemma 3, ChatGPT Plus, and Claude Pro on real previously unseen wound images","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5TUPF2YVEO6A3VAFZKAUUFVQFE","json":"https://pith.science/pith/5TUPF2YVEO6A3VAFZKAUUFVQFE.json","graph_json":"https://pith.science/api/pith-number/5TUPF2YVEO6A3VAFZKAUUFVQFE/graph.json","events_json":"https://pith.science/api/pith-number/5TUPF2YVEO6A3VAFZKAUUFVQFE/events.json","paper":"https://pith.science/paper/5TUPF2YV"},"agent_actions":{"view_html":"https://pith.science/pith/5TUPF2YVEO6A3VAFZKAUUFVQFE","download_json":"https://pith.science/pith/5TUPF2YVEO6A3VAFZKAUUFVQFE.json","view_paper":"https://pith.science/paper/5TUPF2YV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.11200&json=true","fetch_graph":"https://pith.science/api/pith-number/5TUPF2YVEO6A3VAFZKAUUFVQFE/graph.json","fetch_events":"https://pith.science/api/pith-number/5TUPF2YVEO6A3VAFZKAUUFVQFE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5TUPF2YVEO6A3VAFZKAUUFVQFE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5TUPF2YVEO6A3VAFZKAUUFVQFE/action/storage_attestation","attest_author":"https://pith.science/pith/5TUPF2YVEO6A3VAFZKAUUFVQFE/action/author_attestation","sign_citation":"https://pith.science/pith/5TUPF2YVEO6A3VAFZKAUUFVQFE/action/citation_signature","submit_replication":"https://pith.science/pith/5TUPF2YVEO6A3VAFZKAUUFVQFE/action/replication_record"}},"created_at":"2026-07-05T11:39:14.010400+00:00","updated_at":"2026-07-05T11:39:14.010400+00:00"}