GroundAnything: Reconciling Parallel Decoding with Precise Visual Grounding at Flash Speed
作者: Qize Yu, Lianrui Fan, Bowen Ping, Xini Ding, Zetian Song, Junbo Niu, Kaixuan Wang, Tianxing Chen, Yue Chen, Minghua He, Yuran Wang, Jie Huang, Haojun Zhang, Min Chen, Hao Li, Wenxuan Song, Ruihai Wu, Xianming Liu, Shilong Liu, Shuchang Zhou, Ping Luo, Shiyu Huang
分类: cs.CV, cs.AI, cs.LG, cs.RO
发布日期: 2026-09-30
备注: 61 pages, including supplementary material. Project page: https://groundingpi.github.io/groundanything/ Code: https://github.com/groundingpi/GroundAnything Model: https://huggingface.co/GroundingPI/GroundAnything, https://huggingface.co/GroundingPI/GroundAnything-VLM
💡 一句话要点
提出GroundAnything以解决视觉定位速度与精度的矛盾问题
🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)
关键词: 视觉定位 自回归模型 双向扩散 块状去噪 实时系统 性能优化 多任务学习
📋 核心要点
- 现有自回归模型在空间预测中存在序列延迟，导致定位速度慢且效率低下。
- 论文提出通过双向扩散方法来实现并行解码，允许空间假设同时生成并经过迭代去噪进行联合优化。
- 实验结果显示，GroundAnything在多个基准测试中超越了现有模型，尤其在速度和精度上均有显著提升。
📝 摘要（中文）
自回归（AR）基础的定位模型在空间预测时存在序列延迟，并对输出令牌施加因果顺序限制。本文将定位视为视觉证据提取，认为对象、位置和空间关系由图像和查询共同约束，且其依赖关系并不意味着固有的从左到右生成顺序。为此，提出了GroundAnything，一个具有40亿参数的基础模型，通过块状去噪实现快速并行解码与精确定位的结合。训练过程结合了公共数据集的预训练、AR到扩散的直接转换、监督微调和基于GRPO的强化后训练。在30个定位基准测试中，GroundAnything-VLM在同类模型中达到72.42%的新状态，且在快速MTP基础的LocateAnything模型中，平均性能提升至61.75%。
🔬 方法详解
问题定义：本文旨在解决自回归模型在视觉定位任务中因序列生成导致的延迟问题，现有方法在速度和精度上存在矛盾。
核心思路：通过将视觉定位视为视觉证据提取，采用双向扩散方法，允许空间假设并行生成并通过去噪迭代优化，从而提高解码速度和精度。
技术框架：整体架构包括预训练、AR到扩散的转换、联合优化、监督微调和强化后训练等多个阶段，确保模型在不同任务中的适应性和效率。
关键创新：最重要的创新在于引入了块状去噪机制，使得并行解码成为可能，打破了传统自回归模型的因果限制。
关键设计：模型采用了40亿参数的设计，结合了多种损失函数和优化策略，确保在训练过程中能够有效提升模型的性能。具体参数设置和网络结构细节在论文中有详细描述。
🖼️ 关键图片
📊 实验亮点
在实验中，GroundAnything-VLM在30个基准测试中达到72.42%的准确率，超越了同类模型的表现。同时，通过引入熵引导解码，GroundAnything在速度上实现了4.51倍的提升，且在COCO F1mIoU上仅下降0.74个百分点，展现出优异的性能与效率。
🎯 应用场景
该研究的潜在应用领域包括实时视觉定位、自动驾驶、机器人导航等需要快速响应的场景。通过提高视觉定位的速度和精度，GroundAnything能够在延迟敏感的实际系统中发挥重要作用，推动相关技术的进步与应用。未来，该模型有望在更多复杂场景中实现高效的视觉理解与交互。
📄 摘要（原文）
Autoregressive (AR) grounding models serialize spatial predictions, introducing sequential latency and imposing a causal order on output tokens. We view grounding as visual evidence extraction: objects, locations, and spatial relations are jointly constrained by the image and query, yet their dependencies do not imply an intrinsic left-to-right generation order. This distinction makes bidirectional diffusion a natural fit, allowing spatial hypotheses to emerge in parallel and be jointly refined through iterative denoising. We introduce GroundAnything, a 4B-parameter grounding foundation model that reconciles fast parallel decoding with precise localization through blockwise denoising. Training combines grounding pretraining from public datasets and dedicated data engines, direct AR-to-diffusion conversion with joint AR and diffusion objectives, supervised fine-tuning, and GRPO-based reinforcement post-training. Across 30 grounding benchmarks, our autoregressive variant, GroundAnything-VLM, establishes a new overall state of the art among similarly sized models at 72.42%, remaining competitive with GPT-6 Astra (71.35%). With entropy-guided decoding, GroundAnything also surpasses the prior state of the art at this scale, averaging 61.75% versus 53.32% for the fast MTP-based LocateAnything model. We further explore decoding strategies, showing that an optional self-speculative mode achieves a $4.51\times$ speedup over the AR counterpart with a 0.74 percentage-point drop in COCO F1mIoU. Infrastructure experiments show that progressive inference optimizations translate parallel decoding into practical speedups. These support efficient visual grounding in latency-sensitive real-world systems.