helloelwin/topk-ilr

该数据集包含多个配置，每个配置都有相同的特征集，包括问题、弱答案、强答案、真实答案、接受状态、分数、之前准确率和之后准确率。每个配置都有一个训练分割，包含不同数量的字节和示例。数据集的总下载大小和数据集大小也随配置不同而变化。

This dataset contains multiple configurations, each with the same set of features including question, weak answer, strong answer, ground truth answer, acceptance status, score, accuracy before, and accuracy after. Each configuration has a training split with varying numbers of bytes and examples. The total download size and dataset size vary by configuration.

原始信息汇总

数据集概述

数据集配置

gsm8k_0_top0.1

• 特征:
  • question: string
  • weak_answer: string
  • strong_answer: string
  • gt_answer: string
  • accept: bool
  • score: float32
  • acc_before: float64
  • acc_after: float64
• 分割:
  • train:
    • num_bytes: 2514706
    • num_examples: 2414
• 下载大小: 1419014
• 数据集大小: 2514706

gsm8k_1_top0.1

• 特征:
  • question: string
  • weak_answer: string
  • strong_answer: string
  • gt_answer: string
  • accept: bool
  • score: float32
  • acc_before: float64
  • acc_after: float64
• 分割:
  • train:
    • num_bytes: 2377053
    • num_examples: 2294
• 下载大小: 1352228
• 数据集大小: 2377053

gsm8k_2_top0.1

• 特征:
  • question: string
  • weak_answer: string
  • strong_answer: string
  • gt_answer: string
  • accept: bool
  • score: float32
  • acc_before: float64
  • acc_after: float64
• 分割:
  • train:
    • num_bytes: 2309371
    • num_examples: 2220
• 下载大小: 1315307
• 数据集大小: 2309371

gsm8k_3_top0.1

• 特征:
  • question: string
  • weak_answer: string
  • strong_answer: string
  • gt_answer: string
  • accept: bool
  • score: float32
  • acc_before: float64
  • acc_after: float64
• 分割:
  • train:
    • num_bytes: 2235987
    • num_examples: 2133
• 下载大小: 1271049
• 数据集大小: 2235987

数据文件路径

• gsm8k_0_top0.1:
  • train: gsm8k_0_top0.1/train-*
• gsm8k_1_top0.1:
  • train: gsm8k_1_top0.1/train-*
• gsm8k_2_top0.1:
  • train: gsm8k_2_top0.1/train-*
• gsm8k_3_top0.1:
  • train: gsm8k_3_top0.1/train-*