RULER-HQA
Benchmarks
Task NameDataset NameSOTA ResultTrendResults
RULER-HQA 896K
76.56Normalized sub-EM Accuracy
11
RULER-HQA 448K
75Normalized Sub-EM Accuracy
11
RULER-HQA 224K context length
79.69Normalized Sub-EM Accuracy
11
RULER-HQA 112K context length
80.47Normalized Sub-EM Accuracy
11
RULER-HQA 14K context length
0.813Normalized Sub-EM Accuracy
11
RULER-HQA 7K context length
0.82Sub-EM Accuracy
11
RULER-HQA 1.75M context length
78.12Sub-EM Accuracy
5