Search for a command to run...
From Passive Response to Proactive Correction: Enhancing LLM Robustness Against Input Fact Perturbations