Language Models Predict Reading Progress but Fail to Explain Regressions

Researchers from NYU and UMass Amherst found that current language models can accurately predict forward eye movements during human reading, but fail to account for regressions when readers encounter difficulties.

2026-08-19 ~ 2026-08-20 · 2 related posts

• Study: LMs predict forward reading but struggle to explain regressions — tallinzen · 2026-08-19
• LLMs predict forward reading, but fail to explain eye regressions — tallinzen · 2026-08-20