Helpful-Harmless Assistant Dataset (For RLHF)

kaggle2023-11-26 更新2024-03-07 收录

下载链接：

https://www.kaggle.com/datasets/thedevastator/dh-rlhf-helpful-harmless-assistant-dataset

下载链接

链接失效反馈

官方服务：

资源简介：

17k Train, 9000 Test

创建时间：

2023-11-26

Baidicoot/anthropic-helpful-harmless-rlhf

--- dataset_info: features: - name: chosen dtype: string - name: rejected dtype: string - name: prompt dtype: string splits: - name: train num_bytes: 210946983.0 num_ex

Hugging Face2024-05-10 更新60

Anthropic Helpful-Harmless (HH-RLHF), Ultrafeedback (ULTRAFEEDBACK), LMSYS Arena Preferences (LMSYS), PKU-SafeRLHF (SAFERLHF)

本文研究了四个公开的偏好数据集，包括Anthropic Helpful-Harmless (HH-RLHF)、Ultrafeedback (ULTRAFEEDBACK)、LMSYS Arena Preferences (LMSYS)和PKU-SafeRLHF (SAFERLHF)。这些数据集用于训练奖励模型，以支持从人类反馈中进行强化学习（RLHF）。数据集的内容包括提示、两个响应和一个标注，表明

arXiv2024-09-15 更新1050

north/hh-rlhf-harmless

Anthropic的hh-rlhf数据集的内部副本。

Hugging Face2024-11-25 更新120

north/hh-rlhf-helpful-and-harmless

Internal copy of https://huggingface.co/datasets/Anthropic/hh-rlhf.

Hugging Face2024-11-25 更新50

Baidicoot/anthropic-harmless-rlhf

--- dataset_info: features: - name: chosen dtype: string - name: rejected dtype: string - name: prompt dtype: string splits: - name: train num_bytes: 36676366.0 num_exa

Hugging Face2024-05-10 更新40