Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..

VGGT-SLAM: Dense RGB SLAM Optimized on the SL(4) Manifold

Authors: Dominic Maggio, Hyungtae Lim, Luca Carlone

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental As verified by extensive experiments, we demonstrate that VGGT-SLAM achieves improved map quality using long video sequences that are infeasible for VGGT due to its high GPU requirements. Our code is available at https://github.com/MIT-SPARK/VGGT-SLAM. [...] We evaluate VGGT-SLAM on standard RGB SLAM benchmarks to assess both camera pose estimation accuracy and dense mapping quality.
Researcher Affiliation Academia Dominic Maggio Hyungtae Lim Luca Carlone Massachusetts Institute of Technology EMAIL
Pseudocode No The paper describes the methodology in prose and mathematical equations across sections like '4 VGGT-SLAM' and its subsections, but it does not contain any explicit 'Pseudocode' or 'Algorithm' blocks.
Open Source Code Yes Our code is available at https://github.com/MIT-SPARK/VGGT-SLAM.
Open Datasets Yes For evaluation of pose estimation, we employ the 7-Scenes [60] and TUM RGB-D [65] datasets, and report root mean square error (RMSE) of the absolute trajectory error (ATE) using evo [21].
Dataset Splits Yes We evaluate VGGT-SLAM on standard RGB SLAM benchmarks to assess both camera pose estimation accuracy and dense mapping quality. For evaluation of pose estimation, we employ the 7-Scenes [60] and TUM RGB-D [65] datasets, and report root mean square error (RMSE) of the absolute trajectory error (ATE) using evo [21]. [...] Following the protocol of MASt3R-SLAM, we provide dense reconstruction performance on 7-Scenes; see Table 3.
Hardware Specification Yes We use an NVIDIA Ge Force RTX 4090 GPU with AMD Ryzen Threadripper 7960X CPU.
Software Dependencies No The paper mentions 'GTSAM' for backend optimization and 'Levenberg-Marquardt optimizer' as methods used, but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes For parameters, we set wloop = 1, τdisparity = 50 pixels, τinterval = 2, τdesc = 0.8, and τconf = 25%. We also use 300 RANSAC iterations with a threshold of 0.01.