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

Long-Context Language Modeling on RULER (4K to 128K Context Sweep)

90.97Accuracy (8K Context)

FullKV

-2.952421.431345.81570.1987Dec 8, 2025Dec 13, 2025Dec 19, 2025Dec 25, 2025Dec 31, 2025Jan 6, 2026Jan 12, 2026
Updated 4d ago

Evaluation Results

MethodLinks
2025.12
90.97-90.186.1783.0679.6585.99
2025.12
90.3-81.7271.6765.0258.82-0.51
2025.12
90.16-81.4572.5764.0957.03-0.92
2025.12
90.16-81.4572.564.5458.55-0.56
2025.12
90.14-89.9386.283.3876.6-0.74
2025.12
90.1-89.898683.3777.34-0.65
2025.12
90.06-88.8185.7480.474.39-2.11
2025.12
89.9-89.6586.4282.7174.82-1.29
2025.12
89.6-81.9471.566.0960.9174.01
2025.12
88.81-86.0181.1675.6370.48-5.58
2025.12
86.96-79.2568.1965.1257.69-2.57
2025.12
86.53-78.6667.5662.8958.54-3.17
2025.12
85.04-8580.3475.2767.51-7.36
2025.12
84.85-75.7763.2358.1752.22-7.16
2025.12
38.642.733.421.710.9-29.4
2025.12
35.438.733.824.610.7-28.6
2025.12
35.137.43321.210.4-27.4
2025.12
3336.129.117.79-25
2025.12
28.429.917.69.45.9-18.2
2025.12
25.631.6229.55.5-18.8
2025.12
21.17-16.115.92.581-76.64
2025.12
7.77-9.037.17.052.83-67.25
2026.01
3.382.09-----
2026.01
3.151.94-----
2026.01
1.431.731.251.480.790.95-
2026.01
1.171.650.920.650.690.8-
2026.01
1.051.081.041.031.011.01-
2026.01
1.041.071.031.021.011.01-
2026.01
1.031.120.930.830.790.81-
2026.01
0.921.060.820.610.670.79-
2026.01
0.831.190.660.570.550.53-
2026.01
0.741.110.570.490.450.44-
2026.01
0.690.870.610.560.550.54-
2026.01
0.660.850.580.530.510.5-