Skip to content

Latest commit

 

History

History
12 lines (7 loc) · 679 Bytes

README.md

File metadata and controls

12 lines (7 loc) · 679 Bytes

DeltaScore

Welcome to DeltaScore: Fine-Grained Story Evaluation with Perturbations, which has been published in EMNLP2023.

Overview

The goal of this project is to propose a new methodology to conduct fine-grained story evaluation by differentiating perturbations. We have demonstrated that DeltaScore can outperform evaluation methods that use likelihood directly, and you can find the details in our published paper.

Contact

If you have any questions, feedback, feel free to reach out to Zhuohan Xie at zhuohanx@student.unimelb.edu.au. We welcome any inquiries and look forward to hearing from you!