Universal Model Routing for Efficient LLM Inference - Explained Simply | ArXiv Explained