Abstract: Neural implicit representations have had significant impact on simultaneous localization and mapping (SLAM) by enabling robots to build continuous, differentiable, and high-fidelity 3D maps from sensor data. However, as the scale and complexity of the environment grow, neural SLAM approaches face renewed challenges in the back-end optimization process to keep up with runtime requirements and maintain global consistency. We introduce MISO, a hierarchical optimization framework that leverages multiresolution submaps to achieve efficient and scalable neural implicit reconstruction. For local SLAM within each submap, we develop a learned hierarchical optimization scheme that substantially reduces the time needed to optimize the implicit submap features. Further, to correct estimation drift globally, we develop a hierarchical method to align and fuse the multiresolution submap features directly, leading to significant acceleration by avoiding the need to decode full scene geometry. MISO significantly improves computational efficiency and estimation accuracy of neural signed distance function (SDF) SLAM on large-scale real-world benchmarks.