Language Modeling on Wikipedia Welsh (test)
22.1Perplexity (PPL)Specialist
Evaluation results
| Method | Links | |
|---|---|---|
| SpecialistBackbone=Pythia-410M, Seeds=137/20262026.03 | 22.1 | |
| KALAVAI (MoE)Backbone=Pythia-410M, Seeds=137/2026, Routing=perfect routing2026.03 | 22.1 | |
| Base modelBackbone=Pythia-410M, Seeds=137/20262026.03 | 102.7 |