Fig. 3

A The number of tokens our models have trained on increases over time. This line plot shows the number of unique tokens our various machine-learning models see. The x-axis depicts the year, and the y-axis shows the token count. B Earlier years compared to 2010 have greater distances than later years. This confidence interval plot shows the collective distances obtained by sampling 100 tokens present from every year using a single model approach. The x-axis shows a given year, and the y-axis shows the distance metric. C Later years have a lower intra-distance variability compared to the earlier years. This confidence interval plot shows the collective distances obtained by sampling 100 tokens present from every year using our multi-model approach. The x-axis shows a given year, and the y-axis shows the distance metric