Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

UnKEBench

Benchmarks

Task NameDataset NameSOTA ResultTrend
Knowledge EditingUnKEBench Paraphrased questions (Para.)
Bert Score96.6
18
Knowledge EditingUnKEBench Original questions
BERTScore99.86
18
Unstructured Knowledge EditingUnKEBench
Precision (ROUGE-L)39.62
16
Knowledge EditingUnKEBench
Precision52.17
16
Showing 4 of 4 rows