外观
论文arXiv 2506.018442025-06-02
SmolVLA：单卡训练端侧部署的高效视觉语言动作模型
本文由 AI 根据公开内容整理，可能有误，请以原文为准：SmolVLA: A Vision-Language-Action Model for Affordable and Efficient Robotics（arXiv 2506.01844）
核心内容
针对现有视觉语言动作（VLA）模型参数量大、训练成本高且难以在真实场景部署的问题，研究团队推出SmolVLA。该模型专为低成本机器人平台设计，大幅降低训练与推理门槛，同时保持竞争力。评测显示，SmolVLA在仿真及真实机器人基准测试中，性能可比肩参数量大10倍的同类模型，验证了小参数VLA的有效性。
技术架构上，SmolVLA支持单GPU训练及消费级GPU/CPU端侧部署。为解决小模型响应延迟，引入异步推理栈，将感知与动作预测同执行解耦，并通过分块动作生成提升控制频率。项目完全开源，释放了所有代码、预训练权重及基于社区廉价平台采集的训练数据，填补了开源VLA对低成本硬件适配的空白。
关键看点
- 轻量化部署：支持单GPU训练与消费级GPU/CPU推理，显著降低VLA应用门槛
- 异步推理栈：解耦感知预测与动作执行，采用分块生成机制提升控制频率
- 性能对标：小参数量模型在仿真与真机基准中达到10倍大模型同等水平
- 全量开源：开放代码、预训练权重及社区采集数据，适配低成本机器人平台
工程与产业观察
SmolVLA证明了小参数VLA经架构优化可逼近大模型性能，为具身智能端侧部署提供可行路径。其异步推理栈与社区数据集开源，降低了算法工程师在消费级硬件上验证VLA的成本，推动具身大模型从云端算力堆叠向高效端侧工程化落地转型。
英文摘要 (Abstract)
Vision-language models (VLMs) pretrained on large-scale multimodal datasets encode rich visual and linguistic knowledge, making them a strong foundation for robotics. Rather than training robotic policies from scratch, recent approaches adapt VLMs into vision-language-action (VLA) models that enable natural language-driven perception and control. However, existing VLAs are typically massive--often with billions of parameters--leading to high training costs and limited real-world deployability. Moreover, they rely on academic and industrial datasets, overlooking the growing availability of community-collected data from affordable robotic platforms. In this work, we present SmolVLA, a small, efficient, and community-driven VLA that drastically reduces both training and inference costs, while retaining competitive performance. SmolVLA is designed to be trained on a single GPU and deployed on consumer-grade GPUs or even CPUs. To further improve responsiveness, we introduce an asynchronous inference stack decoupling perception and action prediction from action execution, allowing higher control rates with chunked action generation. Despite its compact size, SmolVLA achieves performance comparable to VLAs that are 10x larger. We evaluate SmolVLA on a range of both simulated as well as real-world robotic benchmarks and release all code, pretrained models, and training data.
作者：Mustafa Shukor, Dana Aubakirova, Francesco Capuano, Pepijn Kooijmans, Steven Palma, Adil Zouitine, Michel Aractingi, Caroline Pascal, Martino Russi, Andres Marafioti, Simon Alibert, Matthieu Cord, Tho…
来源
本页事实对应的公开页面，以原文为准。