遇见数据集

teamcore/DPO_Pm3B_U0_beta0.25dpo_proEurus_RM_7bbt_noise_flip_paper0.6_nu0.008

收藏
Hugging Face2025-10-21 更新2025-10-25 收录
官方服务:

资源简介:

该数据集是一个包含源文本、指令、模型和完成情况的文本数据集,用于评估模型在多个方面的性能,如帮助性、诚实性、指令遵循度和真实性。每个完成情况都包含了详细的评分和评分理由。数据集还提供了正确答案和错误答案,以及不同模型的得分信息。

The dataset is a text dataset containing source text, instructions, models, and completions, designed to evaluate model performance in various aspects such as helpfulness, honesty, instruction adherence, and truthfulness. Each completion includes detailed ratings and rating reasons. The dataset also provides correct and incorrect answers, as well as score information for different models.

提供机构:
teamcore
二维码
社区交流群
二维码
科研交流群
商业服务