English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Recent psycholinguistic studies have drawn conflicting conclusions about therelationship between the quality of a language model and the ability of itssurprisal estimates to predict human reading times, which has been speculatedto be due to the large gap in both the amount of training data and modelcapacity across studies. The current work aims to consolidate these findings byevaluating surprisal estimates from Transformer-based language model variantsthat vary systematically in the amount of training data and model capacity ontheir ability to predict human reading times. The results show that surprisalestimates from most variants with contemporary model capacities provide thebest fit after seeing about two billion training tokens, after which they beginto diverge from humanlike expectations. Additionally, newly-trained smallermodel variants reveal a 'tipping point' at convergence, after which thedecrease in language model perplexity begins to result in poorer fits to humanreading times. These results suggest that the massive amount of training datais mainly responsible for the poorer fit achieved by surprisal from largerpre-trained language models, and that a certain degree of model capacity isnecessary for Transformer-based language models to capture humanlikeexpectations.

Transformer-Based Language Model Surprisal Predicts Human Reading Times Best with About Two Billion Training Tokens