Linguistic Probing on BLiMP
64.8PerformanceMasked-Diffusion
Evaluation Results
| Method | Links | |
|---|---|---|
| Masked-DiffusionRepetitions=32, Alpha (α)=02025.12 | 64.8 | |
| DualAlpha (α)=63/64, Data repetitions=12025.12 | 63.7 | |
| DualRepetitions=128, Alpha (α)=1/82025.12 | 63.5 | |
| Masked-DiffusionRepetitions=128, Alpha (α)=0:12025.12 | 63.3 | |
| DualRepetitions=32, Alpha (α)=3/42025.12 | 62.7 | |
| AutoregressiveAlpha (α)=1, Data repetitions=12025.12 | 61.3 | |
| DualAlpha (α)=3/4, Data repetitions=322025.12 | 57.9 | |
| DualAlpha (α)=1/8, Data repetitions=1282025.12 | 56.1 | |
| AutoregressiveAlpha (α)=1, Data repetitions=322025.12 | 53.3 | |
| AutoregressiveAlpha (α)=1, Data repetitions=1282025.12 | 33.2 |