teamcore/DPO_Pm3B_U0_beta0.25sigmoidEurus_RM_7bbt_noise_flip_paper0.6g
收藏Hugging Face2025-10-23 更新2025-10-25 收录
下载链接：
https://hf-mirror.com/datasets/teamcore/DPO_Pm3B_U0_beta0.25sigmoidEurus_RM_7bbt_noise_flip_paper0.6g
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了源文本、指令、模型名称、完成情况（包括帮助性、诚实性、指令遵循性、真实性等多个维度的评分和理由）、批评、自定义系统提示、细粒度分数、模型名称、总分数、原则、响应、正确答案、错误答案、选择和拒绝的答案、Eurus_RM_7b模型的选择和拒绝的分数、响应、生成的奖励分数、选择的奖励分数和GPT评分及反馈。数据集分为默认的一个split，包含1000个示例。
The dataset includes source text, instructions, model names, completions (including ratings and rationales for helpfulness, honesty, instruction following, truthfulness, etc.), critiques, custom system prompts, fine-grained scores, model names, overall scores, principles, responses, correct answers, incorrect answers, chosen and rejected answers, scores for chosen and rejected by Eurus_RM_7b model, responses, generated reward scores, chosen reward scores, and GPT scores and feedback. The dataset is split into a default split containing 1000 examples.
提供机构：
teamcore
