BriefGPT.xyz
Jul, 2020
利用生成模型结合段落检索进行开放域问答
Leveraging Passage Retrieval with Generative Models for Open Domain Question Answering
HTML
PDF
Gautier Izacard, Edouard Grave
TL;DR
研究使用生成模型在开放领域问答中的表现，发现利用文本段落检索可以显著提高性能，证明生成模型擅长聚合和组合多个段落的证据，同时在 Natural Questions 和 TriviaQA 开放式基准测试中取得最好的结果。
Abstract
generative models
for
open domain question answering
have proven to be competitive, without resorting to external knowledge. While promising, this approach requires to use models with billions of parameters, whic
→