AmanPriyanshu/

reasoning-sft-minimax-stratified-kmeans-diverse-reasoning-842K-only

Reasoning SFT dataset generated by MiniMaxAI/MiniMax-M2.5 on prompts from the Stratified K-Means Diverse Reasoning 100K-1M dataset.

Each row has three columns:

input — list of dicts [{"role": "...", "content": "..."}, ...] (conversation turns) response — model-generated response with reasoning block source — task category (math, code, science, chat, safety)

Model:… See the full description on the dataset page: https://huggingface.co/datasets/AmanPriyanshu/reasoning-sft-minimax-stratified-kmeans-diverse-reasoning-842K-only.