<ul><li>Proposed few-shot optimization technique, HED-LM, enhances example selection for sensor-based classification tasks using Large Language Models (LLMs).</li><li>HED-LM combines Euclidean distance-based filtering with contextual relevance scoring by LLMs for selecting examples effectively.</li><li>Study applied HED-LM to fatigue detection using accelerometer data with high variability, achieving a mean macro F1-score of 69.13%.</li><li>Results showed HED-LM outperformed random selection and distance-only filtering, indicating its potential for improving performance in sensor-based learning tasks and broader applications in healthcare and safety scenarios.</li></ul>

Few-Shot Optimization for Sensor Data Using Large Language Models: A Case Study on Fatigue Detection

Discover more