1
The Homogenization Problem in LLMs: Towards Meaningful Diversity in AI Safety
大模型同质化如何威胁AI安全？这篇论文系统剖析了症结与出路。
arXiv:2601.06116v5 Announce Type: replace-cross Abstract: Generative AI models reproduce the human biases in their training data and further amplify t…