LLM-Based NL-to-PDDL Generation with Iterative Repair

This method improves the translation of natural language to PDDL planning problems by using a pipeline of LLM generation, domain-conformance checking, and iterative repair. It addresses faithfulness issues where generated problems are syntactically valid but semantically incorrect relative to the original description.