NARUTO is a neural active reconstruction system that combines a hybrid neural representation with uncertainty learning to enable high-fidelity surface reconstruction. The system uses a multi-resolution hash-grid as the mapping backbone, chosen for its fast convergence and ability to capture high-frequency local features. A key component is the uncertainty learning module, which dynamically quantifies reconstruction uncertainty while actively reconstructing the environment. This module enables a novel uncertainty aggregation strategy for goal searching and efficient path planning. The system autonomously explores by targeting uncertain observations and reconstructs environments with high completeness and fidelity. It also enhances state-of-the-art neural SLAM systems through an active ray sampling strategy. Evaluations in various environments using an indoor scene simulator confirm NARUTO's superior performance and state-of-the-art status in active reconstruction, as evidenced by its results on benchmark datasets like Replica and MP3D. NARUTO is the first neural active reconstruction system capable of functioning in large-scale environments with unrestricted movement. It introduces a hybrid neural representation with a novel uncertainty-aware planning module, enabling 6DoF movement in unrestricted spaces. The system's key contributions include the first neural active reconstruction system with 6DoF movement, an uncertainty learning module for real-time uncertainty quantification, a novel uncertainty-aware planning module, an active ray sampling strategy, and improved reconstruction performance. The system's effectiveness is demonstrated through experiments on Replica and MP3D datasets, showing superior performance in terms of reconstruction accuracy, completeness, and quality. NARUTO's integration of uncertainty learning and active planning enables efficient and accurate 3D reconstruction in complex environments.NARUTO is a neural active reconstruction system that combines a hybrid neural representation with uncertainty learning to enable high-fidelity surface reconstruction. The system uses a multi-resolution hash-grid as the mapping backbone, chosen for its fast convergence and ability to capture high-frequency local features. A key component is the uncertainty learning module, which dynamically quantifies reconstruction uncertainty while actively reconstructing the environment. This module enables a novel uncertainty aggregation strategy for goal searching and efficient path planning. The system autonomously explores by targeting uncertain observations and reconstructs environments with high completeness and fidelity. It also enhances state-of-the-art neural SLAM systems through an active ray sampling strategy. Evaluations in various environments using an indoor scene simulator confirm NARUTO's superior performance and state-of-the-art status in active reconstruction, as evidenced by its results on benchmark datasets like Replica and MP3D. NARUTO is the first neural active reconstruction system capable of functioning in large-scale environments with unrestricted movement. It introduces a hybrid neural representation with a novel uncertainty-aware planning module, enabling 6DoF movement in unrestricted spaces. The system's key contributions include the first neural active reconstruction system with 6DoF movement, an uncertainty learning module for real-time uncertainty quantification, a novel uncertainty-aware planning module, an active ray sampling strategy, and improved reconstruction performance. The system's effectiveness is demonstrated through experiments on Replica and MP3D datasets, showing superior performance in terms of reconstruction accuracy, completeness, and quality. NARUTO's integration of uncertainty learning and active planning enables efficient and accurate 3D reconstruction in complex environments.