--- dataset_info: features: - name: chosen list: - name: content dtype: string - name: role dtype: string - name: rejected list: - name: content dtype: string - name: role dtype: string splits: - name: train num_bytes: 315800721 num_examples: 160800 - name: test num_bytes: 16951579 num_examples: 8552 download_size: 184692118 dataset_size: 332752300 configs: - config_name: default data_files: - split: train path: data/train-* - split: test path: data/test-* --- ## Formatted hh-rlhf This is a version of anthropic [hh-rlhf](https://huggingface.co/datasets/Anthropic/hh-rlhf?row=0) safety dataset, parsed for using with chat templates.