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

RULER-NIAH

Benchmarks

Task NameDataset NameSOTA ResultTrend
Long-context retrievalRULER-NIAH 128k
Accuracy97.2
9
Long-context retrievalRULER-NIAH 64k
Accuracy98.8
9
Long-context retrievalRULER-NIAH 32k
Accuracy100
9
Long-context retrievalRULER-NIAH 16k
Accuracy100
9
Long-context retrievalRULER-NIAH 4k
Accuracy100
9
Showing 5 of 5 rows