Multi-domain instruction dataset covering six specialized areas—coding, STEM, humanities, creative writing, logic puzzles, and data analysis. 200K high-quality samples with focus rotation and anti-generic filtering, designed for supervised fine-tuning.
