ff-anon/feedback-forensics-annotations

Feedback Forensics Annotations This dataset contains the personality annotations from the experiments included in the Feedback Forensics paper. The annotations are provided over pairwise model outputs, typically consisting of a prompt and two responses. Each individual annotation considers as single personality trait (e.g. "confidence"). The annotation indicates when two responses differ with respect to that trait (e.g. one response is more confident), a trait does not apply to… See the full description on the dataset page: https://huggingface.co/datasets/ff-anon/feedback-forensics-annotations.