{"doi":"10.48550/arxiv.2401.08396","title":"Hidden flaws behind expert-level accuracy of multimodal GPT-4 vision in medicine","abstract":"(NEJM) Image Challenges - an imaging quiz designed to test the knowledge and diagnostic capabilities of medical professionals. Evaluation results confirmed that GPT-4V performs comparatively to human physicians regarding multi-choice accuracy (81.6% vs. 77.8%). GPT-4V also performs well in cases where physicians incorrectly answer, with over 78% accuracy. However, we discovered that GPT-4V frequently presents flawed rationales in cases where it makes the correct final choices (35.5%), most prominent in image comprehension (27.2%). Regardless of GPT-4V's high accuracy in multi-choice questions, our findings emphasize the necessity for further in-depth evaluations of its rationales before integrating such multimodal AI models into clinical workflows.","journal":"PubMed","year":2024,"id":487730,"datarank":0.3551662869056973,"base_score":1.0986122886681096,"endowment":1.0986122886681096,"self_citation_contribution":0.16479184330021646,"citation_network_contribution":0.19037444360548084,"self_endowment_contribution":0.16479184330021646,"citer_contribution":0.19037444360548084,"corpus_percentile":null,"corpus_rank":null,"citation_count":2,"citer_count":1,"citers_with_citation_signal":1,"citers_with_endowment":1,"datacite_reuse_total":0,"is_dataset":false,"is_dataset_confidence":0.9502,"is_data_producer":false,"deposit_databanks":null,"is_oa":true,"file_count":0,"downloads":0,"has_version_chain":false,"published_date":"2024-01-01","fair_score":null,"fair_percentile":null,"algorithm_id":"datarank_citation_only_1hop_v6","ranking_scope":"data_only","authors":[{"id":559185,"name":"Fangyuan Chen","orcid":"0000-0002-9891-4753","position":1,"is_corresponding":false},{"id":1194207,"name":"Yiliang Zhou","orcid":"0009-0002-7457-7075","position":2,"is_corresponding":false},{"id":1014293,"name":"Ziyang Xu","orcid":"0009-0007-3228-9821","position":3,"is_corresponding":false},{"id":561419,"name":"Justin M. Cheung","orcid":null,"position":4,"is_corresponding":false},{"id":1201020,"name":"Robert F. Chen","orcid":"0000-0003-0625-9323","position":5,"is_corresponding":false},{"id":104554,"name":"Ronald M. Summers","orcid":"0000-0001-8081-7376","position":6,"is_corresponding":false},{"id":59496,"name":"Justin F. Rousseau","orcid":"0000-0002-2817-9124","position":7,"is_corresponding":false},{"id":649544,"name":"Peiyun Ni","orcid":"0000-0002-9261-0839","position":8,"is_corresponding":false},{"id":1201021,"name":"Marc Landsman","orcid":"0009-0005-8641-0338","position":9,"is_corresponding":false},{"id":349921,"name":"Sally L. Baxter","orcid":"0000-0002-5271-7690","position":10,"is_corresponding":false},{"id":253720,"name":"Subhi J. Al’Aref","orcid":"0000-0002-2258-2035","position":11,"is_corresponding":false},{"id":218325,"name":"Yijia Li","orcid":"0000-0001-5300-0571","position":12,"is_corresponding":false},{"id":905188,"name":"Alexander Chen","orcid":"0000-0002-3183-0088","position":13,"is_corresponding":false},{"id":1201791,"name":"Josef A. Brejt","orcid":null,"position":14,"is_corresponding":false},{"id":264173,"name":"Michael F. Chiang","orcid":"0000-0002-8172-7636","position":15,"is_corresponding":false},{"id":85506,"name":"Yifan Peng","orcid":"0000-0001-9309-8331","position":16,"is_corresponding":false},{"id":45332,"name":"ZHIYONG LU","orcid":"0000-0001-9998-916X","position":17,"is_corresponding":false},{"id":1014353,"name":"Qiao Jin","orcid":"0000-0002-1268-7239","position":0,"is_corresponding":true}],"reference_count":19,"raw_metadata":{"citation_network_status":"fetched"},"created_at":"2026-07-19T02:08:10.215754Z","pmid":"38410646","pmcid":null,"fwci":null,"citation_percentile":null,"influential_citations":0,"oa_status":null,"license":null,"views":0,"total_file_size_bytes":0,"version_count":0,"fair_f":null,"fair_a":null,"fair_i":null,"fair_r":null,"fair_zscore":null,"fair_rationale":null,"fair_model":null,"fair_agent_version":null,"fair_fulltext_source":null,"fair_has_llm":null,"fair_computed_at":null,"clinical_trials":[],"software_tools":[],"db_accessions":[],"linked_datasets":[],"topics":[]}