analiza aiqbench

This study demonstrates that current large language models capture next-word prediction but fail to account for the structural integration effort humans experience during complex sentence processing. The findings highlight a significant gap between probabilistic word forecasting and the cognitive mechanisms required to resolve syntactic ambiguity, challenging existing frameworks in natural language understanding research. This dissociation suggests future AI systems must evolve beyond simple probability estimation to better model human-like comprehension of context-dependent meaning.