DailyLLM: Context-Aware Activity Log Generation Using Multi-Modal Sensors and LLMs

TL;DR

DailyLLM is a novel system that leverages multi-modal smartphone and smartwatch sensors combined with lightweight LLMs to generate rich, accurate, and efficient activity logs across multiple contextual dimensions, advancing user behavior analysis.

Contribution

It introduces the first comprehensive activity log generation system integrating location, motion, environment, and physiology using only common sensors and lightweight LLMs.

Findings

Outperforms state-of-the-art methods in log accuracy and semantic richness.

Achieves 17% higher BERTScore precision with a smaller 1.5B-parameter model.

Delivers nearly 10x faster inference speed on personal devices.

Abstract

Rich and context-aware activity logs facilitate user behavior analysis and health monitoring, making them a key research focus in ubiquitous computing. The remarkable semantic understanding and generation capabilities of Large Language Models (LLMs) have recently created new opportunities for activity log generation. However, existing methods continue to exhibit notable limitations in terms of accuracy, efficiency, and semantic richness. To address these challenges, we propose DailyLLM. To the best of our knowledge, this is the first log generation and summarization system that comprehensively integrates contextual activity information across four dimensions: location, motion, environment, and physiology, using only sensors commonly available on smartphones and smartwatches. To achieve this, DailyLLM introduces a lightweight LLM-based framework that integrates structured prompting with efficient feature extraction to enable high-level activity understanding. Extensive experiments demonstrate that DailyLLM outperforms state-of-the-art (SOTA) log generation methods and can be efficiently deployed on personal computers and Raspberry Pi. Utilizing only a 1.5B-parameter LLM model, DailyLLM achieves a 17% improvement in log generation BERTScore precision compared to the 70B-parameter SOTA baseline, while delivering nearly 10x faster inference speed.