GLM-5.1-Reasoning-1M-Cleaned is a cleaned and reformatted derivative of Kassadin88/GLM-5.1-1000000x. It preserves the original four-subset layout (main, PHD-Science, Multilingual-STEM, Math) while converting every example into a unified SFT-ready schema with explicit conversations, input, output, domain, and meta fields.
This release was prepared from the original dataset published by Kassadin88.
Teacher model in the data: GLM-5.1… See the full description on the dataset page:
https://huggingface.co/datasets/zhangbo2008/GLM-5.1-Reasoning-1M-Cleaned.
