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

> Research article (Findings of the Association for Computational Linguistics: EMNLP 2023, 2023) · cited 11× · AI/ML

**Wikidata**: [openalex:W4389519307](https://www.wikidata.org/wiki/openalex:W4389519307)  
**Source**: https://4ort.xyz/entity/transformer-based-language-model-surprisal-predicts-human-reading-times-best-with-about-two-billion-training-tokens