EMNLP paper with Sebastian Padó: We show that surprisal-based transformer predictions for human reading times diverge for some agreement attraction configurations (ORC) but not others (PP-mod). We discuss theoretical and methodological implications.
arxiv.org/abs/2603.16574
arxiv.org
Diverging Transformer Predictions for Human Sentence Processing: A Comprehensive Analysis of Agreement Attraction Effects
Transformers underlie almost all state-of-the-art language models in computational linguistics, yet their cognitive adequacy as models of human sentence processing remains disputed. In this work, we u...
