Datasets:
metadata
language:
- en
pretty_name: distilabel_math_dpo
size_categories:
- 1K<n<10K
tags:
- pairwise-preference
dataset_info:
features:
- name: prompt
dtype: string
- name: chosen
dtype: string
- name: rejected
dtype: string
- name: chosen_model
dtype: string
- name: rejected_model
dtype: string
- name: chosen_rating
dtype: float64
- name: rejected_rating
dtype: float64
- name: domain
dtype: string
- name: prompt_source
dtype: string
- name: judgement_source
dtype: string
splits:
- name: train
num_examples: 2394
- name: test
num_examples: 24