List your company

Causal Language Modeling on WikiText-103 (val)

72.57Validation PerplexityGEM (N = 2)
72.171674.860877.5580.2392Apr 23, 2026

Evaluation results

MethodLinks
GEM (N = 2)Activation=GEM (N = 2), Throughput (tok/s)=8,603, Standard Deviation=0.162026.0472.574.28471.19
GEM (N = 1)Activation=GEM (N = 1), Throughput (tok/s)=8,604, Standard Deviation=0.132026.0473.324.29480.44
GELUActivation=GELU, Throughput (tok/s)=8,601, Standard Deviation=0.302026.0473.764.3008—
GELU (tanh)Activation=GELU (tanh), Throughput (tok/s)=8,478, Standard Deviation=0.302026.0473.764.30080
ReLUActivation=ReLU, Throughput (tok/s)=8,634, Standard Deviation=0.192026.0477.644.35213.88
SiLU/SwishActivation=SiLU/Swish, Throughput (tok/s)=8,594, Standard Deviation=0.782026.0482.534.41318.77