Answerable-or-Not is an open-source dataset for training efficient deep learning (DL)-based answerability determining prompt filters. The dataset is curated based on the lower level of Do-Not-Answer's hierarchical safety taxonomy. This is a labeled and balanced dataset that comprises of 2440 textual prompts, where each category of the lower level of hierarchical safety taxonomy has 40 textual prompts (20 YES and 20 NO labels). The… See the full description on the dataset page: https://huggingface.co/datasets/kalyannakka/Answerable-or-Not.
