A Statistical Framework of Watermarks for Large Language Models: Pivot, Detection Efficiency and Optimal Rules.

大型語言模型浮水印的統計性架構：樞紐、偵測效率與最適規則