Anthropic Helpful and Harmless
收藏arXiv2025-09-30 收录
下载链接：
https://huggingface.co/datasets/Anthropic/hh-rlhf
下载链接
链接失效反馈官方服务：
资源简介：
该数据集名为“Anthropic”，其中包含了旨在评估模型响应在帮助性和无害性方面的对话。此外，该数据集还用于对Pythia-2.8B模型进行微调。其所涉及的任务是单轮对话。
The dataset named "Anthropic" comprises dialogues designed to assess the helpfulness and harmlessness of model responses. Additionally, this dataset is employed for fine-tuning the Pythia-2.8B model, and the tasks involved are single-turn dialogues.
提供机构：
Anthropic
