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

MAT-PREF

Benchmarks

Task NameDataset NameSOTA ResultTrend
Compositional ReasoningMAT-PREF OOD-property
Logical Question Accuracy60
8
Compositional ReasoningMAT-PREF (OOD-host)
Logical Question Accuracy71.6
8
Compositional ReasoningMAT-PREF (IID)
Logical Question Accuracy65.2
8
Showing 3 of 3 rows