场景即词元：面向通用三维视觉语言理解的多尺度正态分布变换分词器
Scenes as Tokens: Multi-Scale Normal Distributions Transform Tokenizer for General 3D Vision-Language Understanding
把高分辨率点云经多尺度 NDT 表示压缩成 LLM 可消费的整体场景 token，并用同一个解码器同时承担点/框/掩码提示编码与掩码解码，为 3D VLM 提供了一套可复用的「场景分词 + 统一提示接口」范式。
原文依据 a novel three-stage scene tokenization pipeline built upon a Multi-Scale Normal Distributions Transform (NDT) representation
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论近期三维视觉语言模型（VLM）的进展凸显了三维场景理解与推理的巨大潜力。然而，如何有效地将三维场景分词为整体场景词元，并在多样的三维理解任务中复用这些词元，仍然极具挑战性。我们提出 NDTokenizer3D，一个通用型三维 VLM，可执行广泛的三维场景理解任务，同时天然支持人类交互，从而在语言级推理与三维空间理解之间建立桥梁。我们方法的核心是一个新颖的三阶段场景分词流程，它构建于多尺度正态分布变换（NDT）表示之上，并配以多尺度 NDT 解码器（MSDec）。具体而言，NDTokenizer3D 首先从原始高分辨率点云构建多尺度 NDT 表示，同时保留全局上下文与细粒度几何细节。随后，MSDec 逐步融合跨尺度 NDT 特征，生成可被 LLM 端消费的整体场景词元。除分词之外，MSDec 还被复用为人类交互式提示（点、框、掩码）与分割掩码解码的通用接口，从而在单一架构内统一了多样的三维场景理解任务。凭借这种紧凑而统一的设计，NDTokenizer3D 提供了一个细粒度、通用型的三维 VLM，在三维指代分割、三维视觉问答与三维密集描述任务上取得了显著提升。
英文原文摘要 1462 字符
Recent advances in 3D vision-language models (VLMs) highlight a strong potential for 3D scene understanding and reasoning.However, effectively tokenizing 3D scenes into holistic scene tokens, and leveraging these tokens across diverse 3D understanding tasks, remain highly challenging. We present NDTokenizer3D, a generalist 3D VLM that performs a wide range of 3D scene understanding tasks while naturally supporting human interactions, thereby bridging language-level reasoning with 3D spatial understanding. The core of our approach is a novel three-stage scene tokenization pipeline built upon a Multi-Scale Normal Distributions Transform (NDT) representation, paired with a Multi-Scale NDT Decoder (MSDec).Specifically, NDTokenizer3D first constructs a multi-scale NDT representation from raw high-resolution point clouds, preserving both global context and fine-grained geometric details. Next, the MSDec progressively fuses cross-scale NDT features, producing holistic scene tokens consumable by LLM endpoints. Beyond tokenization, MSDec is repurposed as a general interface for human-interactive prompting (points, boxes, masks) and segmentation-mask decoding, unifying diverse 3D scene understanding tasks within a single architecture. With this compact and unified design, NDTokenizer3D offers a fine-grained, general-purpose 3D VLM, achieving remarkable improvements in 3D Referring Segmentation, 3D Visual Question Answering, and 3D Dense Captioning.