Bilingual LLMs Predict Human L2 Reading Dynamics via L1 Transfer
October 23, 2026
Bilingual models that incorporate a reader's native language (L1) more accurately predict eye-tracking metrics during second-language (L2) reading tasks. The predictive advantage is most significant for intermediate learners and is influenced by script overlap between languages.
HOW THIS AFFECTS YOU
●
researcherYou can improve language learning models by incorporating cross-lingual transfer characteristics that mimic human L1-driven biases.