Why are language models less surprised than humans? Testing the Parse Multiplicity Mismatch Hypothesis

Documenti analoghi

• Deconstructing sentence disambiguation by joint latent modeling of reading paradigms: LLM surprisal is not enough
  di: Paape, Dario, et al.
  Pubblicazione: (2026)
• Manipulating language models' training data to study syntactic constraint learning: the case of English passivization
  di: Leong, Cara Su-Yi, et al.
  Pubblicazione: (2024)
• Do Language Models' Words Refer?
  di: Mandelkern, Matthew, et al.
  Pubblicazione: (2023)
• SPAWNing Structural Priming Predictions from a Cognitively Motivated Parser
  di: Prasad, Grusha, et al.
  Pubblicazione: (2024)
• Entailment Semantics Can Be Extracted from an Ideal Language Model
  di: Merrill, William, et al.
  Pubblicazione: (2022)