Megatron-BERT (ensemble)

Reported on 3 benchmarks across 1 task · 1 paper · 3 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Natural Language Processing3 results

Reading ComprehensiononRACE
Accuracy· 2019-09-17
90.9
best: 91.4 (ALBERT (Ensemble))
SOTA
Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism arXiv:1909.08053
Reading ComprehensiononRACE
Accuracy (High)· 2019-09-17
90
best: 92.6 (ALBERTxxlarge+DUMA(ensemble))
SOTA
Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism arXiv:1909.08053
Reading ComprehensiononRACE
Accuracy (Middle)· 2019-09-17
93.1
SOTA
Megatron-LM: Training Multi-Billion Parameter Language Models Using Model Parallelism arXiv:1909.08053