Improving Neural Language Models with a Continuous Cache

Benchmark Model Rank Results
language-modelling-on-wikitext-2Grave et al. (2016) - LSTM + continuous cache pointer#32Test perplexity: 68.9
language-modelling-on-wikitext-2Grave et al. (2016) - LSTM#36Test perplexity: 99.3
language-modelling-on-wikitext-103Neural cache model (size = 2,000)#77Test perplexity: 40.8
language-modelling-on-wikitext-103Neural cache model (size = 100)#78Test perplexity: 44.8
language-modelling-on-wikitext-103LSTM#81Test perplexity: 48.7