BAAI/Infinity-Preference
收藏官方服务:
资源简介:
此版本包含59438个来自Infinity-Instruct指令集的均匀采样指令,适用于每种任务类型。每条指令都附有从Gemma-2-9B-IT中抽样的偏好对。该偏好对由特定于任务的偏好属性权重和ArmoRM进行注释。您还可以使用Infinity-Preference为更多模型生成策略数据。
This version contains 59,438 uniformly sampled instructions from the Infinity-Instruct instruction set, applicable to all task types. Each instruction is accompanied by preference pairs sampled from Gemma-2-9B-IT. These preference pairs are annotated using task-specific preference attribute weights and ArmoRM. You can also use Infinity-Preference to generate policy data for additional models.
提供机构:
BAAI_DATA


