This dataset is a processed and split version of the andyrdt/gpt-oss-20b-rollouts dataset.
Dataset Description
This dataset contains conversation rollouts from GPT models, useful for training and evaluating conversational AI systems. This version excludes the largest subsets (NuminaMath-CoT, apps, BeaverTails, WildChat-1M, combined) to reduce overall size while maintaining diverse coverage.
Dataset Structure… See the full description on the dataset page: https://huggingface.co/datasets/AIGym/gpt-oss-20b-combined.
