This research explores how large language models like ChatGPT-4o can detect speech pathologies through few-shot learning, offering both accuracy and interpretability for clinical applications.

• Uses multimodal LLMs to analyze speech patterns and identify potential disorders
• Employs in-context learning rather than traditional fine-tuning approaches
• Addresses the critical interpretability gap in current automated diagnostic tools
• Provides a potential pathway to more accessible and explainable medical diagnostics

Why it matters: This approach could revolutionize clinical speech assessments by providing cost-effective, accurate, and explainable diagnostic assistance to healthcare providers, potentially improving accessibility to speech pathology screenings.

Exploring In-Context Learning Capabilities of ChatGPT for Pathological Speech Detection