Marko Schmellenkamp

This work presents a tool-supported workflow that works well by reproducing common mistakes in propositional logic modelling that were identified by hand in the literature, and shows that, unlike other algorithmic approaches, the LLM-based approach is suitable for very large sets of data.

LLMs can translate natural language descriptions of formal languages into a new representation, NILE, with high accuracy, enabling automated error explanation in educational settings.