Combining Constraint Reasoning and LLMs for Mixed-Integer Program Explanations

This method integrates constraint reasoning with LLMs to provide contrastive explanations for Mixed-Integer Programming (MIP) optimization decisions. By using constraint reasoning to distill necessary information, the approach overcomes LLM reasoning limitations to ensure trustworthy, human-auditable decision support.