从观察到动作：面向工业场景VLA预训练的基于潜变量动作的基元分割
From Observation to Action: Latent Action-based Primitive Segmentation for VLA Pre-training in Industrial Settings
提出无监督动作基元分割与潜变量动作序列提取的流程，可直接用于从工业视频中自动构建VLA预训练数据，为具身智能提供可扩展的数据组织方案。
原文依据 We present a novel unsupervised framework to unlock vast unlabeled human demonstration data from continuous industrial video streams for Vision-Language-Action (VLA) model pre-training.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论我们提出了一种新颖的无监督框架，用于从连续工业视频流中挖掘大量未标注的人类演示数据，以服务于视觉-语言-动作（VLA）模型的预训练。该方法首先训练一个轻量级运动分词器来编码运动动态，然后采用无监督动作分割器，利用一种新颖的“潜变量动作能量”指标来发现并分割语义连贯的动作基元。该流程输出分割后的视频片段及其对应的潜变量动作序列，提供可直接用于VLA预训练的结构化数据。在公开基准和专有电机装配数据集上的评估表明，该方法能有效分割人类在工作站执行的关键任务。通过视觉语言模型进行的进一步聚类和定量评估证实了所发现动作基元的语义连贯性。据我们所知，这是首个从非结构化工业视频中提取和组织VLA预训练数据的全自动端到端系统，为制造业中具身智能的集成提供了可扩展的解决方案。
英文原文摘要 1135 字符
We present a novel unsupervised framework to unlock vast unlabeled human demonstration data from continuous industrial video streams for Vision-Language-Action (VLA) model pre-training. Our method first trains a lightweight motion tokenizer to encode motion dynamics, then employs an unsupervised action segmenter leveraging a novel "Latent Action Energy" metric to discover and segment semantically coherent action primitives. The pipeline outputs both segmented video clips and their corresponding latent action sequences, providing structured data directly suitable for VLA pre-training. Evaluations on public benchmarks and a proprietary electric motor assembly dataset demonstrate effective segmentation of key tasks performed by humans at workstations. Further clustering and quantitative assessment via a Vision-Language Model confirm the semantic coherence of the discovered action primitives. To our knowledge, this is the first fully automated end-to-end system for extracting and organizing VLA pre-training data from unstructured industrial videos, offering a scalable solution for embodied AI integration in manufacturing.