This repo contains the partial dataset used in "MAmmoTH2: Scaling Instructions from the Web". This partial data is coming mostly from the forums like stackexchange. This subset contains very high-quality data to boost LLM performance through instruction tuning.… See the full description on the dataset page:
https://huggingface.co/datasets/TIGER-Lab/WebInstructSub.