Research experience

Selected research in language-model evaluation and early mental-health forecasting.

01

Co-author · Research article

LLM Performance on Standardized Examinations

Compared ChatGPT-5, Gemini 2.5 Pro, ChatGLM-4.5, and DeepSeek-V3 across mathematics, physics, computer science, and geography. The study combines benchmark datasets with custom exam sets to examine reliability, prompt design, and model-specific error patterns.

Learn more about the researchView the paper ↗

LLM evaluationEducational assessmentPrompt strategy

02

Lead author · Research article

Predictive Analytics for Adolescent Mental Health

Developed a multimodal framework that connects academic pressure, sleep, study hours, performance, and psychological indicators to early mental-health forecasting. Four models were compared, with XGBoost reaching an AUC-ROC of about 0.961.

Learn more about the researchView the paper ↗

Machine learningAdolescent mental healthHuman-centered data