Identifying governing equations from observational data is crucial for understanding nonlinear physical systems but remains challenging due to the risk of overfitting. Here we introduce the Bi-Level ...