PKU-Alignment/self-monitor

Self-Monitor数据集是一个专门为监督微调语言模型而设计的数据集，旨在增强模型的自我监控能力。数据集中的例子旨在教会模型监控自己的输出，并检测潜在的欺骗或不对齐现象。数据集关注于模型需要监控自己的推理过程、检测潜在的欺骗性或不对齐输出、更好地与人类价值观和意图对齐以及提高决策过程的透明度的场景。

The Self-Monitor dataset is curated specifically for supervised fine-tuning of language models to enhance their self-monitoring capabilities. The data focuses on scenarios where models need to monitor their own reasoning processes, detect potentially deceptive or misaligned outputs, develop better alignment with human values and intentions, and improve transparency in their decision-making processes.