metadata
task_categories:
- question-answering
language:
- en
pretty_name: GLUE Diagnostics
size_categories:
- 1K<n<10K
configs:
- config_name: predicate_argument_structure
data_files:
- split: test
path: glue_diagnostics_pas.csv
- config_name: lexical_semantics
data_files:
- split: test
path: glue_diagnostics_lexical_semantics.csv
- config_name: logic
data_files:
- split: test
path: glue_diagnostics_logic.csv
- config_name: world_knowledge
data_files:
- split: test
path: glue_diagnostics_knowledge_world.csv
- config_name: commonsense
data_files:
- split: test
path: glue_diagnostics_knowledge_common_sense.csv
Citation
@inproceedings{wang2019glue,
title={{GLUE}: A Multi-Task Benchmark and Analysis Platform for Natural Language Understanding},
author={Wang, Alex and Singh, Amanpreet and Michael, Julian and Hill, Felix and Levy, Omer and Bowman, Samuel R.},
note={In the Proceedings of ICLR.},
year={2019}
}