metadata
dataset_info:
features:
- name: instruction
list:
- name: role
dtype: string
- name: text
dtype: string
- name: instruction_vi
dtype: string
- name: score_rejected
struct:
- name: PROFANITY
dtype: float64
- name: SEVERE_TOXICITY
dtype: float64
- name: INSULT
dtype: float64
- name: THREAT
dtype: float64
- name: TOXICITY
dtype: float64
- name: IDENTITY_ATTACK
dtype: float64
- name: score_prompt
struct:
- name: PROFANITY
dtype: float64
- name: SEVERE_TOXICITY
dtype: float64
- name: INSULT
dtype: float64
- name: THREAT
dtype: float64
- name: TOXICITY
dtype: float64
- name: IDENTITY_ATTACK
dtype: float64
- name: predict
sequence: string
splits:
- name: train
num_bytes: 3496264
num_examples: 10888
download_size: 1716992
dataset_size: 3496264
configs:
- config_name: default
data_files:
- split: train
path: data/train-*