This dataset contains prompts with specified target word counts for both input prompts and target outputs, designed to test and evaluate language models across different length requirements. Word counts are defined as whitespace-separated tokens, providing a consistent and human-interpretable measure of text length.
These datasets are typically used in performance benchmarking of language models… See the full description on the dataset page: https://huggingface.co/datasets/metrum-ai/manipulation-resistant-prompts-1536-96.
