Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Long-Context Retrieval on RULER
Loading...
80.1
Accuracy
Full attention
72.092
74.171
76.25
78.329
Mar 31, 2026
Accuracy
KV-Budget Reduction
Updated 16d ago
Evaluation Results
Method
Method
Links
Accuracy
KV-Budget Reduction
Full attention
Backbone Model=LLaMA-3...
2026.03
80.1
0
Full attention
Backbone Model=LLaMA-3...
2026.03
79.8
0
Quest
Backbone Model=LLaMA-3...
2026.03
79.8
93
Quest
Backbone Model=LLaMA-3...
2026.03
79.8
97
Quest
Backbone Model=LLaMA-3...
2026.03
79.5
98
MAC-Attention
Backbone Model=LLaMA-3...
2026.03
78.8
95
Quest
Backbone Model=LLaMA-3...
2026.03
78.3
99
MAC-Attention
Backbone Model=LLaMA-3...
2026.03
78
99
Quest
Backbone Model=LLaMA-3...
2026.03
77
93
Quest
Backbone Model=LLaMA-3...
2026.03
75.5
97
Full attention
Backbone Model=Phi-4-Mini
2026.03
74.4
0
Quest
Backbone Model=LLaMA-3...
2026.03
73.4
98
MAC-Attention
Backbone Model=Phi-4-Mini
2026.03
73.1
77
Quest
Backbone Model=LLaMA-3...
2026.03
72.4
99
Feedback
Search any
task
Search any
task