AssemblyBench: Physics-Aware Assembly of Complex Industrial Objects

Introduction Assembling objects from parts requires understanding multimodal instructions, linking them to 3D components, and predicting physically plausible 6-DoF motions for each assembly step. Existing datasets focus on simplified scenarios, overlooking shape complexities and assembly trajectories in industrial assemblies. We introduce AssemblyBench, a synthetic dataset of 2,789 industrial objects with multimodal instruction manuals, corresponding 3D part models, and part assembly trajectories. AssemblyBench is designed to facilitate research that bridges instructional manual understanding and the execution of assembly steps, serving as a benchmark for the development of next-generation assembly algorithms. All necessary data for training, validation, and testing are publicly released. At a Glance - Contents: - Total assemblies: 2,789 - Dataset file size: 2.3 GB (zipped), 4.6 GB (extracted) - Split sizes: - Train: 2,231 (all.train.txt) - Val: 278 (all.val.txt) - Test: 280 (all.test.txt) - Steps (parts) per assembly: - Min: 2 - Max: 20 - Mean: 6.7 Other Resources The code associated with the approach will be released separately on GitHub. Citation If you use the AssemblyBench dataset in your research, please cite our contribution: @inproceedingsLi2026AssemblyBench, author = Li, Danrui and Zhang, Jiahao and Egger, Bernhard and Chatterjee, Moitreya and Lohit, Suhas and Marks, Tim K. and Cherian, Anoop, title = AssemblyBench: Physics-Aware Assembly of Complex Industrial Objects, booktitle = IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), year = 2026, License AssemblyBench dataset extends the Assemble-Them-All dataset, originally released under the MIT License. The original data remains under the MIT License. As the Assemble-Them-All dataset uses assets from the Fusion 360 Gallary Dataset, please refer to the Fusion 360 Gallery Dataset License for legal usage. All new annotations and modifications introduced in our release are licensed under the CC-BY-SA-4.0. SPDX-License-Identifier: CC-BY-SA-4.0 Created by Mitsubishi Electric Research Laboratories (MERL), 2026