Bridging Human and LLM Judgments: Understanding and Narrowing the Gap

This repository contains the code used in

Maia Polo, F., Wang, X., Yurochkin, M., Xu, G., Banerjee, M., & Sun, Y. (2025). Bridging Human and LLM Judgments: Understanding and Narrowing the Gap. arXiv preprint arXiv:2508.12792.

This repository is still under construction, and we will properly document it in the next weeks. In case you want to use Bridge, please check our demo notebooks below.

Quick Start

We have prepared some demos:

• Collecting LLM judgments and converting those in probabilities.
• Detecting and measuring human-LLM gaps.
• Aligning human and LLM judgments.

Installation

To use the code in this repository, clone the repo and create a conda environment using:

conda env create --file=environment.yml
conda activate bridge
pip install jupyter notebook

Citing

@article{polo2025bridging,
  title={Bridging Human and LLM Judgments: Understanding and Narrowing the Gap},
  author={Polo, Felipe Maia and Wang, Xinhe and Yurochkin, Mikhail and Xu, Gongjun and Banerjee, Moulinath and Sun, Yuekai},
  journal={arXiv preprint arXiv:2508.12792},
  year={2025}
}