gililior/

WildFrame

This dataset was originally introduced in the paper WildFrame: Comparing Framing in Humans and LLMs on Naturally Occurring Texts, available on arXiv. Code: https://github.com/SLAB-NLP/WildFrame-Eval

The WildFrame dataset is designed to study the framing effect in large language models (LLMs), in comparison to human behavior. It provides a collection of reframed sentences, with human annotators majority vote for sentiment. This dataset enables… See the full description on the dataset page: https://huggingface.co/datasets/gililior/WildFrame.