laurentiubp/openhermes-scored

该数据集包含多个特征，如消息内容、角色、来源、评分（包括连贯性、复杂性、正确性、帮助性、指令遵循、安全性、冗长性和冗长系数）以及聚合评分。数据集分为训练集和测试集，训练集包含183,043个样本，测试集包含1,849个样本。数据集的下载大小为153,548,192字节，总大小为321,462,037字节。

This dataset includes multiple features such as message content, role, source, scores (including coherence, complexity, correctness, helpfulness, instruction-following, safety, verbosity, and verbosity-coeff), and an aggregate score. The dataset is divided into training and test sets, with the training set containing 183,043 samples and the test set containing 1,849 samples. The download size of the dataset is 153,548,192 bytes, and the total size is 321,462,037 bytes.

原始信息汇总

数据集概述

数据集特征

• messages: 包含以下子特征
  • content: 数据类型为 string
  • role: 数据类型为 string
• source: 数据类型为 string
• scores: 包含以下子特征
  • coherence: 数据类型为 float64
  • complexity: 数据类型为 float64
  • correctness: 数据类型为 float64
  • helpfulness: 数据类型为 float64
  • instruction-following: 数据类型为 float64
  • safety: 数据类型为 float64
  • verbosity: 数据类型为 float64
  • verbosity-coeff: 数据类型为 float64
• agg_score: 数据类型为 float64

数据集分割

• train:
  • 字节数: 318247277.5381899
  • 样本数: 183043
• test:
  • 字节数: 3214759.4618101376
  • 样本数: 1849

数据集大小

• 下载大小: 153548192 字节
• 数据集总大小: 321462037.0 字节

配置

• config_name: default
  • data_files:
    • train: 路径为 data/train-*
    • test: 路径为 data/test-*