WildFrame

该数据集名为WildFrame，包含了1000条文本，这些文本是通过重新构建具有明确情感（正面或负面）的真实世界陈述而创建的，并且收集了人类对情感的评价注释。每个数据集中的陈述都由五位不同的注释者进行标注，以确保对情感转变的评分具有可靠性。该数据集的规模为1000条陈述（其中500条为正面，500条为负面）。该数据集的任务是评估文本情感中的框架效应，通过比较LLM（大型语言模型）的响应与人类行为。

This dataset is named WildFrame. It contains 1000 texts created by reconstructing real-world statements with explicit sentiment (positive or negative), and includes human annotated sentiment evaluations. Each statement in the dataset is annotated by five distinct annotators to ensure the reliability of the scoring for sentiment transitions. The dataset consists of 1000 statements in total, with 500 positive and 500 negative ones respectively. The task of this dataset is to evaluate the framing effect in text sentiment by comparing the responses of Large Language Models (LLMs) with human behaviors.