Safety4M

该数据集是一个针对大型语言模型（LLM）安全的令牌级安全奖励数据集，旨在加速未来在LLM安全性领域的研究。它与Saffron-1模型一同发布，任务是为大型语言模型的安全性提供保障。This dataset is a token-level safety reward dataset targeting the safety of Large Language Models (LLMs), which aims to accelerate future research in the field of LLM safety. It is released alongside the Saffron-1 model, with the primary objective of providing safety guarantees for Large Language Models.