BriefGPT.xyz
Dec, 2019
英语语言最小对比基准: BLiMP
BLiMP: A Benchmark of Linguistic Minimal Pairs for English
HTML
PDF
Alex Warstadt, Alicia Parrish, Haokun Liu, Anhad Mohananey, Wei Peng...
TL;DR
BLiMP是一套挑战集，用于评估语言模型对英语中主要语法现象的理解水平。研究表明，现有模型能够可靠地识别形态对比，但在限定词和否定极性项分布以及提取岛等微妙的语法现象上仍面临挑战。
Abstract
We introduce The Benchmark of Linguistic Minimal Pairs (shortened to
blimp
), a challenge set for evaluating what
language models
(LMs) know about major
→