遇见数据集

cmoyacal/hotel-aug-noninformative-a0p50-seed7

收藏
Hugging Face2026-03-30 更新2026-04-12 收录
官方服务:

资源简介:

--- dataset_info: features: - name: prompt dtype: string - name: chosen dtype: string - name: rejected dtype: string - name: metadata struct: - name: correlation_mode dtype: string - name: is_tie dtype: bool - name: label_matches_utility dtype: bool - name: margin dtype: float64 - name: option_one_utility dtype: float64 - name: option_two_utility dtype: float64 - name: spurious_decorrelation struct: - name: correlation_mode dtype: string - name: spurious_assignment dtype: string - name: strategy dtype: string - name: spurious_tracking struct: - name: both_correlated dtype: bool - name: hotel_a struct: - name: assignment_type dtype: string - name: correlation_applied dtype: bool - name: correlation_direction dtype: string - name: correlation_mode dtype: string - name: random_roll dtype: float64 - name: threshold dtype: float64 - name: utility dtype: float64 - name: was_high_utility dtype: bool - name: hotel_b struct: - name: assignment_type dtype: string - name: correlation_applied dtype: bool - name: correlation_direction dtype: string - name: correlation_mode dtype: string - name: random_roll dtype: float64 - name: threshold dtype: float64 - name: utility dtype: float64 - name: was_high_utility dtype: bool - name: mixed_correlation dtype: bool - name: neither_correlated dtype: bool - name: tie_threshold dtype: float64 - name: true_better dtype: string - name: true_label dtype: string splits: - name: train num_bytes: 23388363 num_examples: 15000 download_size: 3731985 dataset_size: 23388363 configs: - config_name: default data_files: - split: train path: data/train-* ---

数据集信息: 特征: - 字段名:提示词(prompt),数据类型:字符串(string) - 字段名:优选回复(chosen),数据类型:字符串(string) - 字段名:弃选回复(rejected),数据类型:字符串(string) - 字段名:元数据(metadata),数据类型:结构体,包含以下子字段: - 字段名:相关模式(correlation_mode),数据类型:字符串(string) - 字段名:是否平局(is_tie),数据类型:布尔型(bool) - 字段名:标签与效用匹配(label_matches_utility),数据类型:布尔型(bool) - 字段名:边际值(margin),数据类型:64位浮点型(float64) - 字段名:选项一效用(option_one_utility),数据类型:64位浮点型(float64) - 字段名:选项二效用(option_two_utility),数据类型:64位浮点型(float64) - 字段名:虚假去相关(spurious_decorrelation),数据类型:结构体,包含以下子字段: - 字段名:相关模式(correlation_mode),数据类型:字符串(string) - 字段名:虚假分配(spurious_assignment),数据类型:字符串(string) - 字段名:策略(strategy),数据类型:字符串(string) - 字段名:虚假追踪(spurious_tracking),数据类型:结构体,包含以下子字段: - 字段名:双方均相关(both_correlated),数据类型:布尔型(bool) - 字段名:酒店A(hotel_a),数据类型:结构体,包含以下子字段: - 字段名:分配类型(assignment_type),数据类型:字符串(string) - 字段名:已应用相关性(correlation_applied),数据类型:布尔型(bool) - 字段名:相关方向(correlation_direction),数据类型:字符串(string) - 字段名:相关模式(correlation_mode),数据类型:字符串(string) - 字段名:随机滚动值(random_roll),数据类型:64位浮点型(float64) - 字段名:阈值(threshold),数据类型:64位浮点型(float64) - 字段名:效用值(utility),数据类型:64位浮点型(float64) - 字段名:曾为高效用项(was_high_utility),数据类型:布尔型(bool) - 字段名:酒店B(hotel_b),数据类型:结构体,包含以下子字段: - 字段名:分配类型(assignment_type),数据类型:字符串(string) - 字段名:已应用相关性(correlation_applied),数据类型:布尔型(bool) - 字段名:相关方向(correlation_direction),数据类型:字符串(string) - 字段名:相关模式(correlation_mode),数据类型:字符串(string) - 字段名:随机滚动值(random_roll),数据类型:64位浮点型(float64) - 字段名:阈值(threshold),数据类型:64位浮点型(float64) - 字段名:效用值(utility),数据类型:64位浮点型(float64) - 字段名:曾为高效用项(was_high_utility),数据类型:布尔型(bool) - 字段名:混合相关(mixed_correlation),数据类型:布尔型(bool) - 字段名:双方均不相关(neither_correlated),数据类型:布尔型(bool) - 字段名:平局阈值(tie_threshold),数据类型:64位浮点型(float64) - 字段名:更优真实选项(true_better),数据类型:字符串(string) - 字段名:真实标签(true_label),数据类型:字符串(string) 数据划分: - 划分名称:train(训练集),字节大小:23388363,样本数量:15000 下载大小:3731985,数据集总大小:23388363 配置项: - 配置名称:default(默认配置),数据文件: - 数据划分:train,文件路径:data/train-*

提供机构:
cmoyacal
二维码
社区交流群
二维码
科研交流群
商业服务