Reproduce “Leveraging Large Language Models for enhanced personalised user experience in Smart Homes” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jordan Rey-Jouanchicot, André Bottaro, Eric Campo, Jean-Léon Bouraoui, Nadine Vigouroux, Frédéric Vella
This research develops an LLM-centric architecture for smart homes that proactively automates actions by integrating user preferences and contextual data. The system demonstrates improved personalization and adaptability by leveraging natural language representation and efficient prompting strategies, with smaller models like Starling Alpha 7B achieving strong performance with low latency for practical deployment.
Key insights
- Natural language representation of smart home states significantly improves performance, especially for smaller LLMs, leading to a 21.9% average performance increase.
- Directly injecting user preferences via a 'directPref' prompting style is highly effective and efficient, consistently improving performance across models and achieving up to a 52.3% improvement for smaller LLMs.
- Smaller, optimized LLMs can achieve competitive performance with large LLMs while drastically reducing processing times, making them practical for edge deployment in real-time smart home applications.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.