datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
eli5-human-vs-ai
ELI5 Human vs AI (long-form)
This dataset is for training and evaluating AI-writing detectors. It was built
as a clean way to compare known AI text against known human text: every human
answer predates ChatGPT by more than three years, so it is genuinely human by
construction, and every AI answer was written by a named 2026 model, so its origin
is certain too. Most detection datasets have to guess at their labels; this one
does not.
ELI5 answers were chosen because they are… See the full description on the dataset page: https://huggingface.co/datasets/mild-rgb/eli5-human-vs-ai.eli5smollm2_17b_instruct_0531_inf_eli5_product_dpoed_c4_lowq_200m2b_subsample20m_grpo_promptsmollm2_17b_instruct_0531_inf_eli5_softgate_dpoed_c4_lowq_200m2b_subsample20m_grpo_prompt