metadata
language:
- ja
JBLiMP
This is the data from "JBLiMP: Japanese Benchmark of Linguistic Minimal Pairs" (Someya and Oseki, 2023). Only the validated pairs used for benchmarks are included, and only in JSONL format, since it's redundant with the TSV.
For details see the original git repo or the paper.