Character-level Language Modeling on Character-level domain-switching dataset
1.659BPC (all)β-MoE
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| β-MoE2026.05 | 1.659 | 4.01 | 0.604 | 0.996 | 6 | |
| β + Ant MoE2026.05 | 1.681 | 4.19 | 0.858 | 0.997 | 2.4 | |
| Standard MoE2026.05 | 1.78 | 6.559 | 0.421 | 0.559 | 9.6 |