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

Unconditional Generation on FineWeb-Edu Mask source 170M-parameter (train)

8.5Entropy

DFM (Teacher)

7.0447.4227.88.178May 8, 2026
Updated 23d ago

Evaluation Results

MethodLinks
2026.05
8.51,473.3561.21,441
2026.05
8.4761.3307754.9
2026.05
8.4359.3152.7348.9
2026.05
8.3190.387184.2
2026.05
8.21,015.7367.1943.5
2026.05
8.2516.6204.9492.4
2026.05
8.2284.7122.7271.6
2026.05
8.218986.3182.5
2026.05
8.193.645.787.9
2026.05
8109.955.2101.9
2026.05
891.445.484.5
2026.05
7.9188.794.8177.4
2026.05
7.752.43751.3
2026.05
7.7386.7201.2363.4
2026.05
7.661.945.661.1
2026.05
7.491.473.191.3
2026.05
7.1157.4132.2149.8