generalcritiquebearishWithout constraints on the language model, surprisal theory makes no falsifiable predictions about human language processing.Computation and Language29 Jul 2026http://arxiv.org/abs/2607.21574v1