通过用户写作样本预测偏好来对齐LLM
Accommodating human preferences is essential for creating aligned LLM agents that deliver personalized and effective interactions. Recent work has shown the potential for LLMs acting as writing...
本文介绍了PROSE，一种通过用户写作样本提升偏好描述精确度的方法。PROSE通过迭代优化和多样本验证，增强了LLM代理对人类偏好的理解，写作质量比现有方法CIPHER提高了33%。结合ICL，效果再提升9%。
