Reproduce “VS-LLM: Visual-Semantic Depression Assessment based on LLM for Drawing Projection Test” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Meiqi Wu, Yaxuan Kang, Xuchen Li, Shiyu Hu, Xiaotang Chen, Yunfeng Kang, Weiqiang Wang, Kaiqi Huang
VS-LLM, a visual-semantic method, assesses depression by analyzing drawing projection tests with multimodal large language models. This approach from the University of Chinese Academy of Sciences and CASIA achieved 87.8% accuracy on PPAT sketches, demonstrating a 17.6% improvement over traditional psychologist assessments.
Key insights
- Integrating temporal analysis of the drawing process with LLM-generated psychological semantic captions significantly enhances the accuracy of depression assessment from DPTs.
- AI-driven assessment can surpass the accuracy and objectivity of human-expert scoring in interpreting complex psychological data from projective tests.
- Leveraging multimodal LLMs with carefully designed "mental prompts" allows for the extraction of nuanced psychological insights from visual input, such as color usage and space utilization.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.