Break

AI2 Israel, Question Understanding • 2020

Break is a human annotated dataset of natural language questions and their Question Decomposition Meaning Representations (QDMRs). Break consists of 83,978 examples sampled from 10 question answering datasets over text, images and databases.

Download Read Paper View Website View Repo

Leaderboard

Top Public Submissions

Details	Created	GED
1 T5 with QDMR pre-trained decoder Chris Coleman and Alex Reneau, Northwestern University	4/28/2021	23%
2 T5 Baseline Chris Coleman and Alex Reneau, Northwestern University	4/20/2021	24%
3 QDMR-BART + special tokens Chris Coleman and Alex Reneau, Northwestern University	9/20/2020	24%
4 QDMR-BARTv1 Chris Coleman and Alex Reneau, Northwestern University	7/18/2020	23%
5 Curriculum-trained CopyNet Chris Coleman and Alex Reneau, Northwestern University	6/30/2020	27%

View Leaderboard

Authors

Tomer Wolfson, Mor Geva, Ankit Gupta, Matt Gardner, Yoav Goldberg, Daniel Deutch, Jonathan Berant