TurkuNLP/FIN-bench

该数据集是基于FIN-bench的多个零样本任务构成的集合，包括类比、算术、因果、情感、实证判断、常识、对齐、意图识别、误解、释义、句子歧义和相似性抽象等任务。每个任务都包含输入文本、目标文本、多选目标和对应的多选分数。数据集转换为HuggingFace Dataset格式，适用于lm-evaluation-harness工具的使用。

This dataset is a collection of zero-shot tasks based on FIN-bench, including analogy, arithmetic, cause and effect, emotion, empirical judgments, general knowledge, alignment, intent recognition, misconceptions, paraphrase, sentence ambiguity, and similarities abstraction tasks. Each task includes input text, target text, multiple-choice targets, and corresponding multiple-choice scores. The dataset is converted to the HuggingFace Dataset format and is suitable for use with the lm-evaluation-harness tool.