List your company

Language Modeling on TAU2

1.148PerplexityMSA-PT
1.147721.149611.15151.15339Jun 11, 2026

Evaluation results

MethodLinks
MSA-PTTraining Strategy=from-scratch sparse pretraining2026.061.148
MSA-CPTTraining Strategy=sparse continued pretraining2026.061.15
FullTraining Strategy=Full-Attention baseline2026.061.155