AI TrackerFrontier desk · AEST
Back to news
normalaitech

Bilingual Language Models Capture Cross-Linguistic Influence in Second-Language Reading

arXiv cs.CL · 23 Oct 2026, 03:42 AEST

Bilingual language models can effectively capture the influence of a reader's first language on second-language reading dynamics. This study shows that matching the readers' first language improves predictions of reading times in a second language.

Key facts

Read aloud

Bilingual language models can effectively capture the influence of a reader's first language on second-language reading dynamics. This study shows that matching the readers' first language improves predictions of reading times in a second language. While monolingual language model LM surprisal is widely established as a predictor of reading times, it cannot capture the non-native L2 reading dynamics that are shaped by the reader's first language L1. We evaluate bilingual models on eye-tracking data of readers from twelve diverse L1 backgrounds reading English texts. We demonstrate that matching the readers' L1 improves the prediction of L2 reading times. These findings provide evidence that bilingual LMs capture aspects of cross-linguistic influence, and that such influence is present in L2 reading behavior.

Sources in this cluster