Babilong
Benchmarks
TaskDatasetSOTA resultTrendResults
BABILong 16k
29.8Accuracy
72
BABILong 8k
34.7Accuracy
65
BABILong 4k
38.5Accuracy (BABILong 4k)
51
BABILong
14.1Err (2k Context)
16
BABILong OOD
98Performance (16K Context)
10
Babilong 16k context length
58QA1 Accuracy
9
BABILong 8K
57Accuracy
6
BABILong 4K
61.2Accuracy
6
BABILong 2K
65.2Accuracy
6
Babilong 128k context length
38QA1 Score
5
Babilong 64k context length
25QA1 Score
5
BABILong
63.54QA Accuracy (8K Context)
3
BABILong 32K context length
9Accuracy
3
BABILong 16K context length
22.2Needle-in-a-Haystack Accuracy (16K)
3