Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Accuracy on Long-context Understanding on LongBench

103Accuracy

QUOKA

35.81653.25870.788.142Feb 9, 2026
Updated 2d ago

Evaluation Results

MethodLinks
2026.02
103
2026.02
102.8
2026.02
99.8
2026.02
99.5
2026.02
99.2
2026.02
98.6
2026.02
97.7
2026.02
97.2
2026.02
96.6
2026.02
96.6
2026.02
94.7
2026.02
94.5
2026.02
94.5
2026.02
92.9
2026.02
92.2
2026.02
91.8
2026.02
90.1
2026.02
87.5
2026.02
87.2
2026.02
87.2
2026.02
86.9
2026.02
86.8
2026.02
85.6
2026.02
85
2026.02
84.2
2026.02
84.2
2026.02
84.2
2026.02
82.8
2026.02
80.8
2026.02
80.5
2026.02
80.2
2026.02
80.1
2026.02
80
2026.02
78.8
2026.02
78.7
2026.02
78.2
2026.02
78.2
2026.02
77.3
2026.02
76.5
2026.02
75.7
2026.02
75.6
2026.02
75.5
2026.02
73.8
2026.02
72.6
2026.02
72.5
2026.02
72.1
2026.02
70.3
2026.02
68.6
2026.02
68.6
2026.02
67.1
2026.02
66.5
2026.02
66
2026.02
65.9
2026.02
63.6
2026.02
62.2
2026.02
62.2
2026.02
58.9
2026.02
55.6
2026.02
46.1
2026.02
38.4