This research demonstrates how Large Language Models (LLMs) can recognize Activities of Daily Living in smart homes without requiring traditional sensor data training.

• Utilizes LLMs' inherent common-sense knowledge about human activities
• Enables zero-shot recognition of daily activities from sensor data
• Offers a training-free alternative to deep learning methods
• Supports applications in healthcare monitoring, energy management, and safety

For healthcare professionals, this breakthrough means more accessible implementation of activity monitoring systems for patient care and eldercare, with reduced data collection requirements and faster deployment.

Large Language Models are Zero-Shot Recognizers for Activities of Daily Living