AWD-LSTM-DRILL + dynamic eval

Reported on 4 benchmarks across 1 task · 1 paper

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

Medical4 results

Language ModellingonPenn Treebank (Word Level)
Test perplexity· 2019-05-14
49.4
best: 20.5 (GPT-3 (Zero-Shot))
Deep Residual Output Layers for Neural Language Generation arXiv:1905.05513
Language ModellingonPenn Treebank (Word Level)
Validation perplexity· 2019-05-14
49.5
best: 36.1 (BERT-Large-CAS)
Deep Residual Output Layers for Neural Language Generation arXiv:1905.05513
Language ModellingonWikiText-2
Test perplexity· 2019-05-14
42
best: 8.21 (SparseGPT (175B, 50% Sparsity))
Deep Residual Output Layers for Neural Language Generation arXiv:1905.05513
Language ModellingonWikiText-2
Validation perplexity· 2019-05-14
43.9
best: 15.69 (GPT-2 (fine-tuned))
Deep Residual Output Layers for Neural Language Generation arXiv:1905.05513