dongyoung4091/hh-rlhf_with_features_flan_t5_large_max_5_xl_zeroshot
收藏Hugging Face2023-11-14 更新2024-03-04 收录
下载链接:
https://hf-mirror.com/datasets/dongyoung4091/hh-rlhf_with_features_flan_t5_large_max_5_xl_zeroshot
下载链接
链接失效反馈官方服务:
资源简介:
---
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
dataset_info:
features:
- name: chosen
dtype: string
- name: rejected
dtype: string
- name: helpfulness_chosen
dtype: int64
- name: helpfulness_rejected
dtype: int64
- name: specificity_chosen
dtype: int64
- name: specificity_rejected
dtype: int64
- name: intent_chosen
dtype: int64
- name: intent_rejected
dtype: int64
- name: factuality_chosen
dtype: int64
- name: factuality_rejected
dtype: int64
- name: easy-to-understand_chosen
dtype: int64
- name: easy-to-understand_rejected
dtype: int64
- name: relevance_chosen
dtype: int64
- name: relevance_rejected
dtype: int64
- name: readability_chosen
dtype: int64
- name: readability_rejected
dtype: int64
- name: enough-detail_chosen
dtype: int64
- name: enough-detail_rejected
dtype: int64
- name: biased:_chosen
dtype: int64
- name: biased:_rejected
dtype: int64
- name: fail-to-consider-individual-preferences_chosen
dtype: int64
- name: fail-to-consider-individual-preferences_rejected
dtype: int64
- name: repetetive_chosen
dtype: int64
- name: repetetive_rejected
dtype: int64
- name: fail-to-consider-context_chosen
dtype: int64
- name: fail-to-consider-context_rejected
dtype: int64
- name: too-long_chosen
dtype: int64
- name: too-long_rejected
dtype: int64
- name: human
dtype: string
- name: assistant_chosen
dtype: string
- name: assistant_rejected
dtype: string
- name: log_score_chosen
dtype: float64
- name: log_score_rejected
dtype: float64
- name: labels
dtype: string
- name: zeroshot_helpfulness_chosen
dtype: int64
- name: zeroshot_helpfulness_rejected
dtype: int64
- name: zeroshot_specificity_chosen
dtype: int64
- name: zeroshot_specificity_rejected
dtype: int64
- name: zeroshot_intent_chosen
dtype: int64
- name: zeroshot_intent_rejected
dtype: int64
- name: zeroshot_factuality_chosen
dtype: int64
- name: zeroshot_factuality_rejected
dtype: int64
- name: zeroshot_easy-to-understand_chosen
dtype: int64
- name: zeroshot_easy-to-understand_rejected
dtype: int64
- name: zeroshot_relevance_chosen
dtype: int64
- name: zeroshot_relevance_rejected
dtype: int64
- name: zeroshot_readability_chosen
dtype: int64
- name: zeroshot_readability_rejected
dtype: int64
- name: zeroshot_enough-detail_chosen
dtype: int64
- name: zeroshot_enough-detail_rejected
dtype: int64
- name: zeroshot_biased:_chosen
dtype: int64
- name: zeroshot_biased:_rejected
dtype: int64
- name: zeroshot_fail-to-consider-individual-preferences_chosen
dtype: int64
- name: zeroshot_fail-to-consider-individual-preferences_rejected
dtype: int64
- name: zeroshot_repetetive_chosen
dtype: int64
- name: zeroshot_repetetive_rejected
dtype: int64
- name: zeroshot_fail-to-consider-context_chosen
dtype: int64
- name: zeroshot_fail-to-consider-context_rejected
dtype: int64
- name: zeroshot_too-long_chosen
dtype: int64
- name: zeroshot_too-long_rejected
dtype: int64
splits:
- name: train
num_bytes: 16425816
num_examples: 9574
- name: test
num_bytes: 16369741
num_examples: 9574
download_size: 16098559
dataset_size: 32795557
---
# Dataset Card for "hh-rlhf_with_features_flan_t5_large_max_5_xl_zeroshot"
[More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
配置项:
- 配置名称:default
数据文件:
- 训练划分集:路径为 data/train-*
- 测试划分集:路径为 data/test-*
数据集信息:
特征字段:
1. 选中回复(chosen):字符串型
2. 拒选回复(rejected):字符串型
3. 选中回复有用性(helpfulness_chosen):64位整数型
4. 拒选回复有用性(helpfulness_rejected):64位整数型
5. 选中回复具体性(specificity_chosen):64位整数型
6. 拒选回复具体性(specificity_rejected):64位整数型
7. 选中回复意图契合度(intent_chosen):64位整数型
8. 拒选回复意图契合度(intent_rejected):64位整数型
9. 选中回复事实准确性(factuality_chosen):64位整数型
10. 拒选回复事实准确性(factuality_rejected):64位整数型
11. 选中回复易懂性(easy-to-understand_chosen):64位整数型
12. 拒选回复易懂性(easy-to-understand_rejected):64位整数型
13. 选中回复相关性(relevance_chosen):64位整数型
14. 拒选回复相关性(relevance_rejected):64位整数型
15. 选中回复可读性(readability_chosen):64位整数型
16. 拒选回复可读性(readability_rejected):64位整数型
17. 选中回复信息量充足度(enough-detail_chosen):64位整数型
18. 拒选回复信息量充足度(enough-detail_rejected):64位整数型
19. 选中回复偏见性(biased:_chosen):64位整数型
20. 拒选回复偏见性(biased:_rejected):64位整数型
21. 选中回复未考虑个体偏好程度(fail-to-consider-individual-preferences_chosen):64位整数型
22. 拒选回复未考虑个体偏好程度(fail-to-consider-individual-preferences_rejected):64位整数型
23. 选中回复重复冗余度(repetetive_chosen):64位整数型
24. 拒选回复重复冗余度(repetetive_rejected):64位整数型
25. 选中回复未考虑上下文程度(fail-to-consider-context_chosen):64位整数型
26. 拒选回复未考虑上下文程度(fail-to-consider-context_rejected):64位整数型
27. 选中回复过长问题程度(too-long_chosen):64位整数型
28. 拒选回复过长问题程度(too-long_rejected):64位整数型
29. 用户原始输入(human):字符串型
30. 选中的助手回复(assistant_chosen):字符串型
31. 拒选的助手回复(assistant_rejected):字符串型
32. 选中回复对数得分(log_score_chosen):64位浮点数型
33. 拒选回复对数得分(log_score_rejected):64位浮点数型
34. 标签(labels):字符串型
35. 零样本(Zero-shot)选中回复有用性(zeroshot_helpfulness_chosen):64位整数型
36. 零样本(Zero-shot)拒选回复有用性(zeroshot_helpfulness_rejected):64位整数型
37. 零样本(Zero-shot)选中回复具体性(zeroshot_specificity_chosen):64位整数型
38. 零样本(Zero-shot)拒选回复具体性(zeroshot_specificity_rejected):64位整数型
39. 零样本(Zero-shot)选中回复意图契合度(zeroshot_intent_chosen):64位整数型
40. 零样本(Zero-shot)拒选回复意图契合度(zeroshot_intent_rejected):64位整数型
41. 零样本(Zero-shot)选中回复事实准确性(zeroshot_factuality_chosen):64位整数型
42. 零样本(Zero-shot)拒选回复事实准确性(zeroshot_factuality_rejected):64位整数型
43. 零样本(Zero-shot)选中回复易懂性(zeroshot_easy-to-understand_chosen):64位整数型
44. 零样本(Zero-shot)拒选回复易懂性(zeroshot_easy-to-understand_rejected):64位整数型
45. 零样本(Zero-shot)选中回复相关性(zeroshot_relevance_chosen):64位整数型
46. 零样本(Zero-shot)拒选回复相关性(zeroshot_relevance_rejected):64位整数型
47. 零样本(Zero-shot)选中回复可读性(zeroshot_readability_chosen):64位整数型
48. 零样本(Zero-shot)拒选回复可读性(zeroshot_readability_rejected):64位整数型
49. 零样本(Zero-shot)选中回复信息量充足度(zeroshot_enough-detail_chosen):64位整数型
50. 零样本(Zero-shot)拒选回复信息量充足度(zeroshot_enough-detail_rejected):64位整数型
51. 零样本(Zero-shot)选中回复偏见性(zeroshot_biased:_chosen):64位整数型
52. 零样本(Zero-shot)拒选回复偏见性(zeroshot_biased:_rejected):64位整数型
53. 零样本(Zero-shot)选中回复未考虑个体偏好程度(zeroshot_fail-to-consider-individual-preferences_chosen):64位整数型
54. 零样本(Zero-shot)拒选回复未考虑个体偏好程度(zeroshot_fail-to-consider-individual-preferences_rejected):64位整数型
55. 零样本(Zero-shot)选中回复重复冗余度(zeroshot_repetetive_chosen):64位整数型
56. 零样本(Zero-shot)拒选回复重复冗余度(zeroshot_repetetive_rejected):64位整数型
57. 零样本(Zero-shot)选中回复未考虑上下文程度(zeroshot_fail-to-consider-context_chosen):64位整数型
58. 零样本(Zero-shot)拒选回复未考虑上下文程度(zeroshot_fail-to-consider-context_rejected):64位整数型
59. 零样本(Zero-shot)选中回复过长问题程度(zeroshot_too-long_chosen):64位整数型
60. 零样本(Zero-shot)拒选回复过长问题程度(zeroshot_too-long_rejected):64位整数型
划分集信息:
- 训练划分集:字节数为16425816,样本量为9574
- 测试划分集:字节数为16369741,样本量为9574
下载总大小:16098559,数据集总大小:32795557
---
# 「hh-rlhf_with_features_flan_t5_large_max_5_xl_zeroshot」数据集卡片
[更多信息待补充](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
提供机构:
dongyoung4091原始信息汇总
数据集概述
数据集配置
- 配置名称: default
- 数据文件:
- 训练集: data/train-*
- 测试集: data/test-*
数据集信息
- 特征:
- chosen: 类型为字符串
- rejected: 类型为字符串
- helpfulness_chosen: 类型为64位整数
- helpfulness_rejected: 类型为64位整数
- specificity_chosen: 类型为64位整数
- specificity_rejected: 类型为64位整数
- intent_chosen: 类型为64位整数
- intent_rejected: 类型为64位整数
- factuality_chosen: 类型为64位整数
- factuality_rejected: 类型为64位整数
- easy-to-understand_chosen: 类型为64位整数
- easy-to-understand_rejected: 类型为64位整数
- relevance_chosen: 类型为64位整数
- relevance_rejected: 类型为64位整数
- readability_chosen: 类型为64位整数
- readability_rejected: 类型为64位整数
- enough-detail_chosen: 类型为64位整数
- enough-detail_rejected: 类型为64位整数
- biased:_chosen: 类型为64位整数
- biased:_rejected: 类型为64位整数
- fail-to-consider-individual-preferences_chosen: 类型为64位整数
- fail-to-consider-individual-preferences_rejected: 类型为64位整数
- repetetive_chosen: 类型为64位整数
- repetetive_rejected: 类型为64位整数
- fail-to-consider-context_chosen: 类型为64位整数
- fail-to-consider-context_rejected: 类型为64位整数
- too-long_chosen: 类型为64位整数
- too-long_rejected: 类型为64位整数
- human: 类型为字符串
- assistant_chosen: 类型为字符串
- assistant_rejected: 类型为字符串
- log_score_chosen: 类型为64位浮点数
- log_score_rejected: 类型为64位浮点数
- labels: 类型为字符串
- zeroshot_helpfulness_chosen: 类型为64位整数
- zeroshot_helpfulness_rejected: 类型为64位整数
- zeroshot_specificity_chosen: 类型为64位整数
- zeroshot_specificity_rejected: 类型为64位整数
- zeroshot_intent_chosen: 类型为64位整数
- zeroshot_intent_rejected: 类型为64位整数
- zeroshot_factuality_chosen: 类型为64位整数
- zeroshot_factuality_rejected: 类型为64位整数
- zeroshot_easy-to-understand_chosen: 类型为64位整数
- zeroshot_easy-to-understand_rejected: 类型为64位整数
- zeroshot_relevance_chosen: 类型为64位整数
- zeroshot_relevance_rejected: 类型为64位整数
- zeroshot_readability_chosen: 类型为64位整数
- zeroshot_readability_rejected: 类型为64位整数
- zeroshot_enough-detail_chosen: 类型为64位整数
- zeroshot_enough-detail_rejected: 类型为64位整数
- zeroshot_biased:_chosen: 类型为64位整数
- zeroshot_biased:_rejected: 类型为64位整数
- zeroshot_fail-to-consider-individual-preferences_chosen: 类型为64位整数
- zeroshot_fail-to-consider-individual-preferences_rejected: 类型为64位整数
- zeroshot_repetetive_chosen: 类型为64位整数
- zeroshot_repetetive_rejected: 类型为64位整数
- zeroshot_fail-to-consider-context_chosen: 类型为64位整数
- zeroshot_fail-to-consider-context_rejected: 类型为64位整数
- zeroshot_too-long_chosen: 类型为64位整数
- zeroshot_too-long_rejected: 类型为64位整数
数据集分割
- 训练集:
- 字节数: 16425816
- 样本数: 9574
- 测试集:
- 字节数: 16369741
- 样本数: 9574
数据集大小
- 下载大小: 16098559
- 数据集大小: 32795557



