Hill Climbing is prone to falling into local optima, and the quality of the evaluation suite directly determines whether the optimization direction is correct; if scenario coverage is insufficient, a high pass rate does not equal high-quality performance in real-world scenarios.