Case-report–based Retrieval-Augmented Generation of Large Language Models for Identification of Malignant Melanocytic Lesions by Anatomic Site

Supplementary Figure 1. Schematic Representation of Study Design and Three-Phase Workflow for Evaluating Multimodal LLM Performance in Dermoscopic Melanocytic Lesion Classification Supplementary Table 1. Case reports (n=648) with DOI Supplementary Table 2. Exact System and User Prompts for Phases 1 and 3 Supplementary Table 3. Raw study data generated using get-5-mini (n = 9,999), including ISIC image ID, histopathologically confirmed diagnosis, anatomical region, Phase 1 Diagnosis, and Phase 3 Diagnosis Supplementary Table 4. Raw study data generated using gemini-3.0-flash (n = 9,999), including ISIC image ID, histopathologically confirmed diagnosis, anatomical region, Phase 1 Diagnosis, and Phase 3 Diagnosis