arcee-ai/OpenHermes2.5-dpo-binarized-alpha

---
dataset_info:
features:
- name: prompt
dtype: string
- 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: 32434312
num_examples: 8813
- name: test
num_bytes: 3489590
num_examples: 980
download_size: 20557079
dataset_size: 35923902
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
language:
- en
tags:
- dpo
- orpo
- distilabel
- synthetic
size_categories:
- 1K<n<10K
---
**OpenHermes-2.5-DPO-binarized-alpha**
A DPO dataset built with distilabel atop the awesome `OpenHermes-2.5` dataset.

This is an alpha version with a small sample to collect feedback from the community. It follows a fully OSS approach, using PairRM for preference selection instead of OpenAI models.

Reference: https://huggingface.co/datasets/argilla/OpenHermes2.5-dpo-binarized-alpha?row=0

原始信息汇总

数据集概述

数据集信息

• 名称: OpenHermes-2.5-DPO-binarized-alpha
• 描述: 基于OpenHermes-2.5数据集，使用distilabel构建的DPO数据集的alpha版本。

数据集特征

• prompt: 字符串类型
• chosen:
  • content: 字符串类型
  • role: 字符串类型
• rejected:
  • content: 字符串类型
  • role: 字符串类型

数据集分割

• train:
  • 示例数量: 8813
  • 字节数: 32434312
• test:
  • 示例数量: 980
  • 字节数: 3489590

数据集大小

• 下载大小: 20557079字节
• 数据集大小: 35923902字节

配置

• 默认配置:
  • 训练数据路径: data/train-*
  • 测试数据路径: data/test-*

语言

• en

标签

• dpo
• orpo
• distilabel
• synthetic

大小类别

• 1K<n<10K