The Stacked Multi-Model Reasoning (SMMR) framework leverages multiple LLMs and specialized smaller models as collaborative experts to improve mental health assessments from long-form text.

• Organizes AI models in layers where early layers handle discrete subtasks and later layers integrate findings
• Creates a cooperative architecture where specialized models work together rather than relying on a single large model
• Demonstrates superior performance on clinical mental health datasets including depression screening
• Reduces LLM hallucination issues when processing extensive, domain-specific content

This research matters for healthcare by providing more reliable automated assessment tools that can help address the global shortage of mental health practitioners while maintaining clinical accuracy.

A Layered Multi-Expert Framework for Long-Context Mental Health Assessments