This is a in-house development version of KLUE Topic Classification benchmark, as the test split is not released by the KLUE team.

We randomly split the original validation set (9,107 instances) into in-house validation set (5,107 instances) and the in-house test set (4,000 instances).

Models trained or fine-tuned on naver-clova-conversation/klue-tc-dev-tsv

None yet