Using LLMs for Late Multimodal Sensor Fusion for Activity Recognition | Dark Hacker News