LLM Evaluation for Mental-Health Counseling

Evaluated proprietary and open-source LLMs (ChatGPT, Claude, LLaMA, Deepseek) on counseling capabilities using classification and generation tasks with comprehensive metrics. Developed detailed annotation guidelines and analyzed inter-rater reliability to systematically assess therapeutic response quality across models. The study examines empathy, safety, clinical appropriateness, and actionability of AI-generated counseling responses, providing insights into the readiness of LLMs for mental health applications.