Reproduce “Bayesian Preference Elicitation with Language Models” 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
Kunal Handa, Yarin Gal, Ellie Pavlick, Noah Goodman, Jacob Andreas, Alex Tamkin, Belinda Z. Li
OPEN learns readers’ news preferences by using a language model to describe candidate features and phrase pairwise comparisons as ordinary questions. A Bayesian model selects questions to reduce uncertainty and updates preference weights from responses. In tests on real New York Times article pairs, OPEN prediction outperformed language-model prediction on the same questions, within this fixed-feature GPT-4-based study.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.