Research in federated learning (FL) for the Internet of Vehicles (IoV) focuses on enabling collaborative, privacy-preserving, and intelligent data analytics across distributed vehicular networks without sharing raw data. Recent studies propose hierarchical and asynchronous FL architectures to handle the high mobility, dynamic topology, and heterogeneous computing capabilities of vehicles. Edge-assisted and blockchain-integrated FL frameworks are being developed to ensure secure model aggregation, trust management, and resistance against poisoning or inference attacks. Deep reinforcement learning and transfer learning are also employed to optimize resource allocation, model convergence, and communication efficiency during training. These advancements enhance real-time decision-making for applications such as traffic prediction, autonomous driving, and vehicular intrusion detection, thereby improving safety, privacy, and scalability in next-generation IoV systems.