Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Long-context language modeling on RULER Sequence length = 16k
Loading...
100
S-NIAH Component 1
HqeKV
95
97.5
100
102.5
May 8, 2026
S-NIAH Component 1
S-NIAH Component 2
S-NIAH Component 3
MK-NIAH Component 1
MK-NIAH Component 2
MK-NIAH Component 3
MQ-NIAH Score
MV-NIAH Score
VT Score
CWE Score
FWE Score
Average Score
Updated 22d ago
Evaluation Results
Method
Method
Links
S-NIAH Component 1
S-NIAH Component 2
S-NIAH Component 3
MK-NIAH Component 1
MK-NIAH Component 2
MK-NIAH Component 3
MQ-NIAH Score
MV-NIAH Score
VT Score
CWE Score
FWE Score
Average Score
HqeKV
Backbone=LLaMA-3.1-8B-...
2026.05
100
99.4
99
99.6
99.8
86.4
96.7
95.1
96.2
26.8
95
90.4
RDKV
Backbone=LLaMA-3.1-8B-...
2026.05
100
99.8
100
99.6
99.6
86.2
99.4
98.5
99.2
76.7
93.2
95.7
Feedback
Search any
task
Search any
task
