datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
Rotten_Tomatoesrotten_tomatoessycophancy_rotten_tomatoes
Sycophancy Rotten Tomatoes Dataset
The generated dataset includes a text (chat between a human and an assistant), the sycophancy of the exchange, and additional information.
Dataset Structure
The dataset is structured as follows:
text: The generated prompt text of the chat between the human and the assistant.
assistant_opinion: The assistant's opinion, converted to a label (i.e. its final answer.
human_opinion: The human's opinion, converted to a label.
sycophancy: A… See the full description on the dataset page: https://huggingface.co/datasets/romaingrx/sycophancy_rotten_tomatoes.rotten-tomatoes-gm-datasetrotten_tomatoes_noisyThis dataset is the rotten_tomatoes dataset from huggingface, but with Laplacian noise and randomness added to secure it against adversary models.
It was used for a course project.