Abstract: Differentiable programming (DP) is a paradigm that enables programs to be automatically differentiated by computers, effectively combining the interpretability of traditional mathematical operations with the learning capabilities of neural networks. By integrating automatic differentiation into numerical optimization, DP substantially reduces computational overhead and improves optimization efficiency compared to conventional derivative-free optimization approaches which are difficult to derive derivatives manually. Consequently, this methodology has attracted considerable attention in recent years. With the ongoing scaling of quantum computing systems, the challenges in quantum design automation (QDA) are growing in complexity, and the pressing need for efficient numerical optimization methods further underscores the importance of DP. This article examines recent typical applications of DP across the device, control, and circuit levels of QDA. These implementations enable efficient computation of derivatives with respect to diverse parameters, thereby accelerating the design workflow. As can be seen from this review, DP has been already widely used in QDA and holds significant potential to become a standard paradigm, playing a pivotal role in the future advancement of quantum computing.

Key words: quantum computing, quantum design automation, differentiable programming, automatic differentiation