Your submission is now in Draft mode.

Once it's ready, please submit your draft for review by our team of Community Moderators. Thank you!

Submit Essay

Once you submit your essay, you can no longer edit it.


This content now needs to be approved by community moderators.


This essay was submitted and is waiting for review.

WikiText-103 SOTA Language Modelling

AI Progress Essay Contest

Fan Graph

Fan Graph currently only supports question groups with continuous numeric sub questions! Unsupported configuration detected, you are on your own.


Language modelling has been applied to a wide range of applications and domains with great success. To name a few, automatic speech recognition, machine translation, spelling correction, touchscreen “soft” keyboards and many natural language processing applications depend on the quality of language models.

The WikiText language modelling dataset is a collection of over 100 million tokens extracted from the set of verified Good and Featured articles on Wikipedia. These articles have been reviewed by humans and are considered well written, factually accurate, broad in coverage, neutral in point of view, and stable.

As of December 2020, the state-of-the-art model for is kNN-LM (Khandelwal et al. 2020), which achieves a perplexity of 15.79 on the WikiTex-103 test set.

An excellent reference for tracking state-of-the-art models is PapersWithCode, which tracks performance data of ML models.

What will be the state-of-the-art language modelling performance (in perplexity) on WikiText-103 by the following dates?

The sub-questions below will resolve as the lowest level of perplexity achieved by any language model on WikiText-103's test set before 11:59PM GMT on the date in question. Qualifying models need to be trained on only the WikiText-103's training set—no extra training data may be used.

Performance figures may be taken from e-prints, conference papers, peer-reviewed articles, and blog articles by reputable AI labs (including the associated code repositories). Published performance figures must be available before the date of the subquestion.

In case the relevant performance figure is given as a confidence interval, the median value will be used to resolve the question.

Make a Prediction


Note: this question resolved before its original close time. All of your predictions came after the resolution, so you did not gain (or lose) any points for it.

Note: this question resolved before its original close time. You earned points up until the question resolution, but not afterwards.

Current points depend on your prediction, the community's prediction, and the result. Your total earned points are averaged over the lifetime of the question, so predict early to get as many points as possible! See the FAQ.