The field of visual-inertial navigation systems is witnessing significant advancements, driven by the need for more accurate, efficient, and robust state estimation methods. Recent developments are focused on addressing the long-standing inconsistency issue in Visual-Inertial Navigation Systems (VINS), improving the accuracy and efficiency of filter-based and optimization-based approaches, and exploring new paradigms such as autoregressive proprioceptive odometry and dual-agent reinforcement learning. Notable papers in this area include: Unobservable Subspace Evolution and Alignment for Consistent Visual-Inertial Navigation, which proposes a novel analysis framework and solution paradigm to eliminate inconsistency. SP-VINS: A Hybrid Stereo Visual Inertial Navigation System based on Implicit Environmental Map, which achieves high computational efficiency while maintaining long-term high-accuracy localization performance. AutoOdom: Learning Auto-regressive Proprioceptive Odometry for Legged Locomotion, which overcomes challenges in legged robot navigation through an innovative two-stage training paradigm. An Efficient Closed-Form Solution to Full Visual-Inertial State Initialization, which presents a closed-form initialization method that recovers the full visual-inertial state without nonlinear optimization. Metric, inertially aligned monocular state estimation via kinetodynamic priors, which enables robust and accurate pose estimation on non-rigid platforms. Dual-Agent Reinforcement Learning for Adaptive and Cost-Aware Visual-Inertial Odometry, which reduces the computational load of optimization-based methods through lightweight reinforcement learning agents. Dual Preintegration for Relative State Estimation, which proposes a novel observation integrating IMU preintegration from both platforms to enable efficient relinearization.