jmajkutewicz/PKU-SafeRLHF-binarized

Dataset Summary This is a binarized version of the PKU-SafeRLHF dataset. It was converted to a format suitable for DPO alignment training by selecting responses with lower severity as preferred responses. Please see the original PKU-SafeRLHF dataset for full dataset details: https://huggingface.co/datasets/PKU-Alignment/PKU-SafeRLHF