metadata
dataset_info:
features:
- name: feedback
dtype: string
- name: generations
dtype: string
- name: input
dtype: string
- name: labelling_model
dtype: string
- name: labelling_prompt
dtype: string
- name: raw_labelling_response
dtype: string
- name: critique
dtype: 'null'
- name: score
dtype: 'null'
splits:
- name: train
num_bytes: 6039711
num_examples: 1000
download_size: 2645043
dataset_size: 6039711
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
Used UltraCM to feedback on feedback-collection dataset. So now we have a reference and generated feedback.