Knowledge Distillation for LLM-Based Human Activity Recognition in Homes

2026-01-12

AI Summary

This study addresses the high computational cost of deploying large language models (LLMs) for human activity recognition (HAR) in home environments. It presents the first systematic evaluation of LLMs of varying scales on HAR tasks and introduces a knowledge distillation–based model compression approach. By transferring the reasoning capabilities of a large LLM to a lightweight student model, the method achieves comparable accuracy to the best-performing large model while retaining only 1/50th of its original parameters. Experimental results on two mainstream HAR datasets demonstrate that the distilled compact model maintains near-optimal performance, offering an efficient pathway for deploying LLM-driven HAR systems in resource-constrained settings.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: (Large) Language Models
• Computer Vision: Large Vision Models

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Human Activity Recognition (HAR) is a central problem for context-aware applications, especially for smart homes and assisted living. A few very recent studies have shown that Large Language Models (LLMs) can be used for HAR at home, reaching high performance and addressing key challenges. In this paper, we provide new experimental results regarding the use of LLMs for HAR, on two state-of-the-art datasets. More specifically, we show how recognition performance evolves depending on the size of the LLM used. Moreover, we experiment on the use of knowledge distillation techniques to fine-tune smaller LLMs with HAR reasoning examples generated by larger LLMs. We show that such fine-tuned models can perform almost as well as the largest LLMs, while having 50 times less parameters.

Problem

• Human Activity Recognition
• Large Language Models
• Knowledge Distillation
• Smart Homes
• Model Compression

Innovation

• Knowledge Distillation
• Large Language Models
• Human Activity Recognition
• Model Compression
• Smart Homes