research
∙
09/06/2021
Finding needles in a haystack: Sampling Structurally-diverse Training Sets from Synthetic Data for Compositional Generalization
Modern semantic parsers suffer from two principal limitations. First, tr...
share
research
∙
10/12/2020
