Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Long-context retrieval and aggregation on RULER 4k

94.73Average Accuracy

NestedKV

16.126836.533456.9477.3466May 26, 2026
Updated 7d ago

Evaluation Results

MethodLinks
2026.05
94.73
2026.05
94.52
2026.05
93.83
2026.05
93.6
2026.05
93.24
2026.05
91.8
2026.05
90.4
2026.05
90.16
2026.05
89.12
2026.05
88.89
2026.05
88.69
2026.05
85.48
2026.05
85.02
2026.05
84
2026.05
82.32
2026.05
80.66
2026.05
80.61
2026.05
80.59
2026.05
79.48
2026.05
79.32
2026.05
79.11
2026.05
77.78
2026.05
77.72
2026.05
77.21
2026.05
76.9
2026.05
75.5
2026.05
75.37
2026.05
75.37
2026.05
75.06
2026.05
73.84
2026.05
73.16
2026.05
72.39
2026.05
72.28
2026.05
70.09
2026.05
68.74
2026.05
66.26
2026.05
65.53
2026.05
65.5
2026.05
63.03
2026.05
61.48
2026.05
61.34
2026.05
60.97
2026.05
60.53
2026.05
60.35
2026.05
60.22
2026.05
59.53
2026.05
58.23
2026.05
56.7
2026.05
56.55
2026.05
56.33
2026.05
52.64
2026.05
50.26
2026.05
49.73
2026.05
43.13
2026.05
42.9
2026.05
42.8
2026.05
42.24
2026.05
40.48
2026.05
40.18
2026.05
39.95
2026.05
37.53
2026.05
34.45
2026.05
33.09
2026.05
32.09
2026.05
29.49
2026.05
28.42
2026.05
28.37
2026.05
28.19
2026.05
25.72
2026.05
24.76
2026.05
22.64
2026.05
22.18
2026.05
21.48
2026.05
20.78
2026.05
20.67
2026.05
19.15