ACL Paper: Existing LLMs Struggle to Accurately Assess Text Readability

windx0303 · x · 2026-07-30

A new paper accepted to ACL reveals that existing automatic readability assessment methods, including LLMs, are poor predictors of actual human reading ease.

Using eye-tracking-based cognitive evaluation, the authors found that these automated metrics significantly diverge from real human cognitive reading experiences.

Original post →

More from Research

Research channel →