This dataset contains project-authored MariChatmen training data from the
May 2026 experiment. It is published so the released adapters and experiment
write-up can be inspected without uploading external-derived transformed data
as if it were project-owned.
MariChatmen is a fictional Sevillian assistant that answers in informal
Andaluh-style written Andalusian Spanish. The data is synthetic,
hand-authored, or hand-authored/template-expanded project data.… See the full description on the dataset page:
https://huggingface.co/datasets/MariChatmen/MariChatmen-Project-Data.
