SupraLabs releases reasoning-corpus-4K-5M-v1 dataset

The dataset contains 5 million samples for training small language models on reasoning tasks. It includes repo_id, question, answer, and reasoning traces.