nyu-mll/

blimp

	Dataset Card for "blimp"




	
	Dataset Summary

BLiMP is a challenge set for evaluating what language models (LMs) know about major grammatical phenomena in English. BLiMP consists of 67 sub-datasets, each containing 1000 minimal pairs isolating specific contrasts in syntax, morphology, or semantics. The data is automatically generated according to expert-crafted grammars.

More Information Needed

More Information… See the full description on the dataset page: https://huggingface.co/datasets/nyu-mll/blimp.