# ArtGen: Conditional Generative Modeling of Articulated Objects in Arbitrary Part-Level States

Source: [ArtGen: Conditional Generative Modeling of Articulated Objects in Arbitrary Part-Level States](<https://www.sota2.com/research/paper/6940e76c456146936284f5df>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-12-17T05:03:45.767Z

Authors: Haowen Wang, Xiaoping Yuan, Fugang Zhang, Rui Jian, Yuanwei Zhu, Xiuquan Qiao, Yakun Huang

Published: 2025-12-13T17:00:03.000Z

## Abstract

Generating articulated assets is crucial for robotics, digital twins, and embodied intelligence. Existing generative models often rely on single-view inputs representing closed states, resulting in ambiguous or unrealistic kinematic structures due to the entanglement between geometric shape and joint dynamics. To address these challenges, we introduce ArtGen, a conditional diffusion-based framework capable of generating articulated 3D objects with accurate geometry and coherent kinematics from single-view images or text descriptions at arbitrary part-level states. Specifically, ArtGen employs cross-state Monte Carlo sampling to explicitly enforce global kinematic consistency, reducing structural-motion entanglement. Additionally, we integrate a Chain-of-Thought reasoning module to infer robust structural priors, such as part semantics, joint types, and connectivity, guiding a sparse-expert Diffusion Transformer to specialize in diverse kinematic interactions. Furthermore, a compositional 3D-VAE latent prior enhanced with local-global attention effectively captures fine-grained geometry and global part-level relationships. Extensive experiments on the PartNet-Mobility benchmark demonstrate that ArtGen significantly outperforms state-of-the-art methods.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [3D Articulated Object Generation on PartNet-Mobility (Geo/Kin/Pref Metrics)](<https://www.sota2.com/research/sota/3d-articulated-object-generation-on-partnet-mobility-geo-kin-pref-metrics.md>): 3D Articulated Object Generation; PartNet-Mobility; Geometric Score: 4.05.

- [Articulated Object Generation (Part Retrieval Shape) on PartNet-Mobility (test)](<https://www.sota2.com/research/sota/articulated-object-generation-part-retrieval-shape-on-partnet-mobility-test.md>): Articulated Object Generation (Part Retrieval Shape); PartNet-Mobility (test); POR: 0.189.

- [3D Articulated Object Generation on PartNet-Mobility](<https://www.sota2.com/research/sota/3d-articulated-object-generation-on-partnet-mobility.md>): 3D Articulated Object Generation; PartNet-Mobility; POR: 0.694.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2512.12395>)

[PDF](<https://arxiv.org/pdf/2512.12395.pdf>)