en license: openrail task_categories:
text-generation tags:
instruction-tuning
identity
safety
alignment
reasoning
pico-1b size_categories:
50K<n<100K
A high-quality, high-variability instruction dataset for PiCo-1B (ArcOffical/PiCo-1B), a ~1.46B parameter dense decoder-only transformer trained from scratch by the Arc Develop Team.
This dataset teaches PiCo-1B who it is, what it can do, what it should do, what it… See the full description on the dataset page:
https://huggingface.co/datasets/ArcOffical/PiCo-DATA-general.
