This research challenges the "bigger is better" paradigm by demonstrating that smaller neural language models can more effectively detect thought disorders in schizophrenia patients than large language models.

• Superior performance: Small neural language models outperform larger models in detecting disorganized thinking patterns
• Clinical practicality: Smaller models address privacy concerns, reduce computational/financial costs, and offer greater transparency
• Practical applications: Potential for more accessible and efficient mental health screening tools in clinical settings
• Paradigm shift: Demonstrates that model size isn't always correlated with effectiveness for specialized tasks

For medical professionals, this research offers a path toward more accessible, efficient, and privacy-preserving tools for mental health assessment that could be deployed in resource-limited settings.

Bigger But Not Better: Small Neural Language Models Outperform Large Language Models in Detection of Thought Disorder