Large Language Model (LLM) surprisal is increasingly used as a proxy for human sentence processing, yet whether model expectations reflect the cognitive factors governing real-time human behavior remains uncertain. We examine Korean subject–verb honorific agreement in subject- and object-control constructions, where human eye-tracking data reveal a prominence-driven asymmetry. Evaluating three Korean-capable models, we find that surprisal was sensitive to honorific-feature mismatches, yielding attraction-like effects, but was not modulated by distractor prominence. Although surprisal selectively improved reading-time predictions, this reflected item-level predictive overlap rather than prominence sensitivity. GPT models thus capture surface regularities but do not reproduce the construction-sensitive, prominence-driven asymmetry of human dependency resolution, suggesting that LLMs diverge from humans precisely where processing depends on working-memory constraints rather than distributional expectation.

This study examines how Transformer language models utilise lexico-phrasal information to interpret the polysemy of the Korean adverbial postposition -(u)lo. We analysed the attention weights of both a Korean pre-trained BERT model and a fine-tuned version. Results show a general reduction in attention weights following fine-tuning, alongside changes in the lexico-phrasal information used, depending on the specific function of -(u)lo. These findings suggest that, while fine-tuning broadly affects a model’s syntactic sensitivity, it may also alter its capacity to leverage lexico-phrasal features according to the function of the target word.

Postpositions, which are characterized as multiple form-function associations and thus polysemous, pose a challenge to automatic identification of their usage. Several studies have used contextualized word-embedding models to reveal the functions of Korean postpositions. Despite the superior classification performance of previous studies, the particular reason how these models resolve the polysemy of Korean postpositions is not enough clear. To add more interpretation, for this reason, we devised a classification model by employing two transformer-architecture models—BERT and GPT-2—and introduces a computational simulation that interactively demonstrates how these transformer-architecture models simulate human interpretation of word-level polysemy involving Korean adverbial postpositions -ey, -eyse, and -(u)lo. Results reveal that (i) the BERT model performs better than the GPT-2 model to classify the intended function of postpositions, (ii) there is an inverse relationship between the classification accuracy and the number of functions that each postposition manifests, (iii) model performance is affected by the corpus size of each function, (iv) the models’ performance gradually improves as the epoch proceeds, and (vi) the models are affected by the scarcity of input and/or semantic closeness between the items.