PL-Mix: A Balanced Polish-Language Dataset for Prompt Harmfulness Detection
Dataset Summary
PL-Mix is a balanced Polish-language dataset designed for training and evaluating prompt-level harmfulness classifiers.
It contains 1,040 prompts, evenly split between harmful (520) and unharmful (520) samples, with an 80/20 stratified train–test split.
The dataset was created to address the lack of publicly available, balanced, and linguistically diverse Polish resources for… See the full description on the dataset page: https://huggingface.co/datasets/mi-crow-team/PL-Mix.
