arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.03533cs.CV
收藏

PnP-U3D: 无缝接入3D框架,连接自回归与扩散以实现统一的理解与生成

PnP-U3D: Plug-and-Play 3D Framework Bridging Autoregression and Diffusion for Unified Understanding and Generation

Shanghai Artificial Intelligence Laboratory(上海人工智能实验室)

Yongwei Chen, Tianyi Wei, Yushi Lan, Zhaoyang Lyu, Shangchen Zhou, Xudong Xu, Xingang Pan

AI总结:

PnP-U3D提出结合自回归与扩散的统一3D理解和生成框架,通过轻量级Transformer实现跨模态信息交换,提升3D生成与编辑性能。

摘要:

大规模多模态模型的快速发展激发了构建能够结合理解和生成的统一框架的努力。尽管此类范式在2D领域已取得显著成功,但将其扩展到3D仍鲜有探索。现有的尝试将3D任务统一到单一的自回归(AR)范式下,导致性能下降,因为强制信号量化和训练成本过高。我们的关键见解是,核心挑战不在于强制统一的自回归范式,而在于在生成和理解之间实现有效的信息交互,同时尽可能不损害其固有能力,并利用预训练模型降低训练成本。基于这一观点,我们提出了首个统一的3D理解和生成框架,结合了自回归与扩散。具体来说,我们采用自回归的下一个标记预测范式用于3D理解,采用连续扩散范式用于3D生成。一个轻量级的Transformer桥接大型语言模型的特征空间和3D扩散模型的条件空间,使有效的跨模态信息交换成为可能,同时保持由独立模型学习的先验知识。大量实验表明,我们的框架在多种3D理解和生成基准上均取得最佳性能,并在3D编辑任务中也表现出色。这些结果凸显了统一AR+扩散模型作为构建更通用3D智能的有前景方向的潜力。

英文摘要:

The rapid progress of large multimodal models has inspired efforts toward unified frameworks that couple understanding and generation. While such paradigms have shown remarkable success in 2D, extending them to 3D remains largely underexplored. Existing attempts to unify 3D tasks under a single autoregressive (AR) paradigm lead to significant performance degradation due to forced signal quantization and prohibitive training cost. Our key insight is that the essential challenge lies not in enforcing a unified autoregressive paradigm, but in enabling effective information interaction between generation and understanding while minimally compromising their inherent capabilities and leveraging pretrained models to reduce training cost. Guided by this perspective, we present the first unified framework for 3D understanding and generation that combines autoregression with diffusion. Specifically, we adopt an autoregressive next-token prediction paradigm for 3D understanding, and a continuous diffusion paradigm for 3D generation. A lightweight transformer bridges the feature space of large language models and the conditional space of 3D diffusion models, enabling effective cross-modal information exchange while preserving the priors learned by standalone models. Extensive experiments demonstrate that our framework achieves state-of-the-art performance across diverse 3D understanding and generation benchmarks, while also excelling in 3D editing tasks. These results highlight the potential of unified AR+diffusion models as a promising direction for building more general-purpose 3D intelligence.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑