Eloquent/preference_prediction
收藏Hugging Face2025-03-10 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/Eloquent/preference_prediction
下载链接
链接失效反馈官方服务：
资源简介：
偏好预测数据集，用于评估系统在生成性大型语言模型(LLM)输出方面的预测能力，并根据相关性、自然性、真实性、安全性和整体质量五个标准进行解释。数据集由Toloka AI和奥斯陆大学的语言技术团队策划，包含验证集和测试集，支持细粒度的域外评估。
Preference Prediction Dataset, designed to evaluate the prediction capability of systems on generative large language model (LLM) outputs and explain according to five criteria: relevance, naturalness, truthfulness, safety, and overall quality. Curated by Toloka AI and the Language Technology Group at the University of Oslo, it includes validation and test sets for fine-grained out-of-domain evaluation.
提供机构：
Eloquent
