This repo contains experiment and demo code that leverages GPT-3 for handling ambiguous user commands to smart assistants. It recieves vague instructions like "get ready for a party" and produces a set of changes to smart home device states in response.
./demo/
contains the Python notebook for the demo application
./experiments.ipynb
contains the experiments from the paper
./results/study.ipynb
contains experiment data post-processing code
./results/markdown/
contains GPT-3 responses from the evaluations in human-readable Markdown format
@misc{king2023get,
title={"Get ready for a party": Exploring smarter smart spaces with help from large language models},
author={Evan King and Haoxiang Yu and Sangsu Lee and Christine Julien},
year={2023},
eprint={2303.14143},
archivePrefix={arXiv},
primaryClass={cs.HC}
}