Transformers over-extend what humans underlearn: the case of Spanish L-shaped morphome
Researchers investigated whether transformers can learn the Spanish L-shaped morphome, an irregular morphological pattern, by training models on varying frequencies of irregular verbs. The study found that while transformers could learn the pattern, their generalization differed qualitatively from human behavior. Specifically, models preferred irregular forms more as their proportion in training increased and showed sensitivity to phonological similarity, unlike humans who favored regular inflections. AI
IMPACT Reveals differences in how AI models generalize linguistic patterns compared to humans, suggesting potential areas for model improvement.