ibm-research/popqa-tp

Dataset Card for "popqa-tp" Dataset Summary PopQA-TP (PopQA Templated Paraphrases) is a dataset derived from PopQA (https://huggingface.co/datasets/akariasai/PopQA), created for the paper "Predicting Question-Answering Performance of Large Language Models through Semantic Consistency". PopQA-TP takes each question in PopQA and paraphrases it using each of several manually-created templates specific to each question category. The paper investigates the… See the full description on the dataset page: https://huggingface.co/datasets/ibm-research/popqa-tp.