List your company

Language Modeling on WikiText-103 5.0% corruption (test)

43.511Perplexity (PPL)Transformer-XL
42.8089647.5477352.286557.02527Mar 31, 2026

Evaluation results

MethodLinks
Transformer-XLModel size (Parameters)=Base (151M), Attention=QUEST2026.0343.511
Transformer-XLModel size (Parameters)=Base (151M), Attention=Standard2026.0344.168
TransformerModel size (Parameters)=Medium (90M), Attention=QUEST2026.0354.531
TransformerModel size (Parameters)=Medium (90M), Attention=QKNorm-DS2026.0354.899
TransformerModel size (Parameters)=Medium (90M), Attention=Standard2026.0355.012
TransformerModel size (Parameters)=Medium (90M), Attention=QNorm2026.0355.451
TransformerModel size (Parameters)=Small (44M), Attention=Standard2026.0360.944
TransformerModel size (Parameters)=Small (44M), Attention=QUEST2026.0361.062