RLHF-And-Friends/Human-vs-Shapa-2x
收藏Hugging Face2025-03-27 更新2025-04-26 收录
下载链接:
https://hf-mirror.com/datasets/RLHF-And-Friends/Human-vs-Shapa-2x
下载链接
链接失效反馈官方服务:
资源简介:
这个数据集包含了来自RLHF-And-Friends/tldr-sft测试分割的人类完成的补全,以及borisshapa/ppo-2x-mistral-7b-smallsft-tldr补全。数据集中的prompt列包含了同时给人类和模型的提示。使用的模型是ppo-2x-mistral-7b-smallsft-tldr。原始数据集,包括提示和人类补全,来源于RLHF-And-Friends/tldr-sft。ppo-2x-mistral-7b-smallsft-tldr模型在gpt-4o-mini意见基础上,相对于人类的胜率是55.4%。
This dataset contains human completions from the RLHF-And-Friends/tldr-sft test split and borisshapa/ppo-2x-mistral-7b-smallsft-tldr completions. The prompt column in the dataset includes prompts given to both humans and models. The model used is ppo-2x-mistral-7b-smallsft-tldr. The original dataset with prompts and human completions is from RLHF-And-Friends/tldr-sft. The winrate of the ppo-2x-mistral-7b-smallsft-tldr model over humans, based on gpt-4o-mini opinion, is 55.4%.
提供机构:
RLHF-And-Friends



