Hierarchical Transformers Are More Efficient Language Models

Open paper
Benchmark Model Rank Results
image-generation-on-imagenet-32x32Hourglass#15bpd: 3.74
language-modelling-on-enwiki8Hourglass#16Bit per Character (BPC): 0.997
image-generation-on-imagenet-64x64Hourglass#39Bits per dim: 3.44

Rank counts only results with a code link.