BriefGPT.xyz
Nov, 2024
利用似然性作为检索增强生成的性能评估工具
Likelihood as a Performance Gauge for Retrieval-Augmented Generation
HTML
PDF
Tianyu Liu, Jirui Qi, Paul He, Arianna Bisazza, Mrinmaya Sachan...
TL;DR
本研究解决了检索增强生成中,大语言模型受文档检索顺序影响的问题,并通过深入分析填补了这一领域的空白。我们提出似然性作为语言模型性能的有效评估工具,通过实验证明了问题的似然性与回答准确性之间的相关性,并提出两种利用问题似然性优化提示选择与构建的方法,显著提升了生成效果。
Abstract
Recent work finds that
Retrieval-augmented generation
with large
Language models
is prone to be influenced by the order of retrieved documents in the context. However, the lack of in-depth analysis limits the use
→