With the transition of urban rail transit toward networked operations, the scientific planning of transit networks is critical for ensuring operational efficiency and service quality. This systematically the research progress on optimization models and methodologies for urban rail transit networks, based on relevant studies published from 2015 to 2025. Current rail transit network optimization models suffer from idealized assumptions and insufficient characterization of practical constraints. Moreover, the three mainstream solution approaches—exact algorithms, intelligent heuristic algorithms, and deep reinforcement learning—each have respective drawbacks in computation efficiency, stability, generalization capacity and other aspects. The paper begins by establishing a taxonomy of transit network planning problems from three perspectives: planning scenarios, modeling formulations, and solution method. Then, in terms of optimization models, the evolution from single-objective to multi-objective formulations is reviewed, highlighting the development of multi-dimensional objective systems involving efficiency, cost, equity, and resilience. This paper further summarizes the main advances in initial network generation models and existing network expansion optimization models. On the algorithmic side, the paper covers exact methods, metaheuristic algorithms, surrogate-assisted optimization, and the emerging deep reinforcement learning (DRL) paradigm, with particular attention to studies that formulate network design as a Markov Decision Process (MDP). It is concluded that the existing optimization models generally have structural defects and deviate from engineering practice. They fail to fully characterize practical constraints such as transfer layouts, capacity limits and time-varying passenger flows, which greatly reduces their practical application value. Five research directions are therefore proposed to improve relevant research: constructing a hybrid optimization framework for large-scale networks, establishing an integrated supply-demand coordination model, developing collaborative optimization theories for multi-mode rail transit systems, forming a low-carbon sustainable optimization system, and building a digital twin-driven dynamic decision-making model.