List your company

Language Modeling on Clustered Pre-training Data Cluster 10

17.783Perplexity (PPL)Split MoE
17.7722417.8448717.917517.99013Mar 19, 2026

Evaluation results

MethodLinks
Split MoEModel Variant=Split MoE2026.0317.783
BaseModel Variant=Base2026.0318.052