Gravity-Aware Hierarchical Routing for Lightweight SensorLLM on Human Activity Recognition
Researchers have developed a novel gravity-aware hierarchical routing head to improve the performance of lightweight sensor-based language models for human activity recognition. This method addresses a failure mode where compressing models like TinyLlama degrades the discrimination of static activities. By extracting statistical cues related to posture and gravity, the system adaptively combines static and full experts, significantly boosting performance on static classes with minimal parameter overhead. AI
IMPACT Enhances the accuracy of AI models for human activity recognition, particularly for static poses.