List your company

Multi-task Language Understanding on MMLU (Accuracy and vNMSE)

73.04AccuracyBF16
71.53271.923572.31572.7065Feb 9, 2026

Evaluation results

MethodLinks
BF16Communication Topology=butterfly all-reduce, Model=LLaMA 1B2026.0273.040
DynamiQCommunication Topology=butterfly all-reduce, Model=LLaMA 1B2026.0273.040.0007
MXFP8Communication Topology=butterfly all-reduce, Model=LLaMA 1B2026.0272.860.002
MXFP6Communication Topology=butterfly all-reduce, Model=LLaMA 1B2026.0272.460.0201
MXFP4Communication Topology=butterfly all-reduce, Model=LLaMA 1B2026.0271.590.1706