The Pupil Becomes the Master: Eye-Tracking Feedback for Tuning LLMs
Abstract
Large language models often require alignment with explicit human preferences, which can be sparse and costly. We propose a framework to leverage eye-tracking data as an implicit feedback signal to tune LLMs for controlled sentiment generation using Direct Preference Optimization. Our study demonstrates that eye-tracking feedback can be a valuable signal for tuning LLMs. This motivates future research to investigate the impact of eye-tracking feedback on various tasks, highlighting the potential of integrating eye-tracking data with LLMs to improve their performance and alignment with human preferences.
Cite
Text
Kiegeland et al. "The Pupil Becomes the Master: Eye-Tracking Feedback for Tuning LLMs." ICML 2024 Workshops: LLMs_and_Cognition, 2024.Markdown
[Kiegeland et al. "The Pupil Becomes the Master: Eye-Tracking Feedback for Tuning LLMs." ICML 2024 Workshops: LLMs_and_Cognition, 2024.](https://mlanthology.org/icmlw/2024/kiegeland2024icmlw-pupil/)BibTeX
@inproceedings{kiegeland2024icmlw-pupil,
title = {{The Pupil Becomes the Master: Eye-Tracking Feedback for Tuning LLMs}},
author = {Kiegeland, Samuel and Reich, David Robert and Cotterell, Ryan and Jäger, Lena Ann and Wilcox, Ethan},
booktitle = {ICML 2024 Workshops: LLMs_and_Cognition},
year = {2024},
url = {https://mlanthology.org/icmlw/2024/kiegeland2024icmlw-pupil/}
}