This dataset corresponds to the paper: CSEval: Towards Automated, Multi-Dimensional, and Reference-Free Counterspeech Evaluation using Auto-Calibrated LLMs (NAACL'25).

	Data Schema

Column Description

uuid Unique identifier for the instance.

hatespeech The input toxic/hateful text.

gold_counterspeech Reference counterspeech used for benchmarking.

predicted_counterspeech The response generated by an LLM.

csType The "type" or "strategy" of counterspeech (e.g.… See the full description on the dataset page: https://huggingface.co/datasets/Aswini123/CSEval.