Evaluate the model

VLM finetuning

Skinalor Agentica is an agent-based dermatology system that explicitly separates clinical decision-making from model inference through structured, orchestrated agents.

Skinalor benchmarks Context vs Prompt Engineering on multimodal dermatology (MedGemma-4B-IT, Lingshu-7B, MedVLM-R1). Tasks: diagnosis/attributes/VQA. Compare zero/few-shot prompts vs RAG contexts; …

This project mainly provides: construction and validation of a RAG system, RAG-based retrieval, and performance evaluation of model baselines and RAG-augmented models.

mental VLM/LLM

test my github