List your company

Language Modeling on FineWeb-Edu 10B (held-out last unused bin)

15.34PerplexityGDN
15.013217.219119.42521.6309May 11, 2026

Evaluation results

MethodLinks
GDNLayers=10, State dim=768, Mem. cells=1.97M (distributed)2026.0515.34
Transformer++Layers=12, State dim=768, Mem. cells=— (KV)2026.0516.05
GPN+MLayers=2, State dim=1920, Mem. cells=360K (centralized)2026.0516.95
GPN+MLayers=1, State dim=2496, Mem. cells=492K (centralized)2026.0518.06
GPN (no memory)Layers=1, State dim=3184, Mem. cells=02026.0523.51