6 April 2026 in IEEE Transactions on Mobile Computing
Large Artificial Intelligence Models (LAMs) have delivered impressive capabilities, but deploying and fine-tuning them on resource-limited mobile edge devices remains challenging due to data privacy concerns, limited computation and memory resources, and prohibitive communication overhead. This paper proposes a novel framework, named Quantized Split Federated Fine-Tuning Large AI Model (SFLAM). By splitting the model across edge devices and an edge server, SFLAM assigns only lightweight computation to devices and offloads the remaining training to the server, substantially reducing the device-side memory footprint while keeping raw data local. However, high-dimensional intermediate activations impose a heavy uplink burden. SFLAM addresses this through activation quantization and derives a convergence bound showing that the upper bound decreases with increasing quantization bit-width. We further formulate an accuracy–energy efficiency objective jointly optimize transmit power, bandwidth allocation, and quantization bit-width to improve training efficiency over wireless links. Simulations under heterogeneous data and wireless conditions show that SFLAM improves training efficiency and scalability over baselines.
26 March 2026 in IEEE Transactions on Image Processing
3D scene CAD recomposition aims to reconstruct a given scene by retrieving and assembling CAD models from a database, so as to accurately simulate the geometric properties and spatial arrangement of the original environment. Recent methods learn this task through training on limited scan-to-CAD annotation data, which hinders their generalization to diverse real-world scenes. In this paper, we propose POSITION, an open-world 3D scene CAD recomposition method to construct the 3D scene with CADs retrieved from an open-set database. POSITION is designed following a divide-and-conquer strategy. Firstly, we extract open-world multi-modal object representations from a captured 3D scene. Secondly, on top of the representations, we propose a coarse-to-fine retrieval method to retrieve CADs that are visually, geometrically and semantically match real objects. Thirdly, we present a physically plausible pose alignment method to adjust retrieved CAD models to maintain consistent geometry and layout with the observation. By decomposing the problem into well-defined subtasks, our approach achieves generalization across various scene types and scalable CAD databases without retraining or fine-tuning. Our approach demonstrates superior CAD recomposition performance on both the Scan2CAD and diverse real-world 3D scene datasets. Our project page: https://yangrongkun.github.io/position/.