A benchmark suite for testing how well AI understands language.
GLUE is like an obstacle course in gym class. The AI must hop over sentence traps, not just sprint.
It compares NLU skill across models. You often see it in papers and leaderboards, especially around BERT.
NLU
GLUE tests NLU with several small language tasks.
BERT
BERT used GLUE scores to help make pretraining famous.
MMLU
MMLU later took the baton for broad model testing.
Leaderboard
Leaderboards use GLUE scores to compare models side by side.