This dataset contains model-generated content that might be offensive. All examples are included for research purposes only and do not reflect or support any opinion.
ConvinceQA accompanies "Prior Beliefs Prejudice LLM-as-Judge: Evidence from Persuasion Evaluation" (ACL 2026 Findings), which shows that LLMs used as judges systematically conflate their trained beliefs with rhetorical quality when rating persuasiveness.… See the full description on the dataset page:
https://huggingface.co/datasets/PardisSzah/ConvinceQA.
