Transformer-Based Language Model Surprisal Predicts Human Reading Times Best with About Two Billion Training Tokens
This addresses a conflict in psycholinguistics about model quality and human reading prediction, with incremental insights into training data effects.
The study investigated how the amount of training data and model capacity in Transformer-based language models affect their surprisal estimates' ability to predict human reading times, finding that best fits occur after about two billion training tokens, beyond which performance diverges from human expectations.
Recent psycholinguistic studies have drawn conflicting conclusions about the relationship between the quality of a language model and the ability of its surprisal estimates to predict human reading times, which has been speculated to be due to the large gap in both the amount of training data and model capacity across studies. The current work aims to consolidate these findings by evaluating surprisal estimates from Transformer-based language model variants that vary systematically in the amount of training data and model capacity on their ability to predict human reading times. The results show that surprisal estimates from most variants with contemporary model capacities provide the best fit after seeing about two billion training tokens, after which they begin to diverge from humanlike expectations. Additionally, newly-trained smaller model variants reveal a 'tipping point' at convergence, after which the decrease in language model perplexity begins to result in poorer fits to human reading times. These results suggest that the massive amount of training data is mainly responsible for the poorer fit achieved by surprisal from larger pre-trained language models, and that a certain degree of model capacity is necessary for Transformer-based language models to capture humanlike expectations.