Skip to content

论文arXiv 2611.1044312 天前

SPARK-VLA:53M参数实现无3D输入VLA空间理解蒸馏 ​

本文由 AI 根据公开内容整理,可能有误,请以原文为准:Spatial Representation Distillation and Knowledge Routing for Vision-Language-Action Models(arXiv 2611.10443)

核心内容 ​

针对现有视觉-语言-动作模型(VLA)依赖显式几何输入或单视觉通路导致语义丢失的问题,研究团队提出SPARK-VLA框架。该方法采用“保留-特化-路由”策略,在不引入外部3D传感器或专家模型推理的前提下,赋予VLA三维物理结构理解能力。实验显示,基于0.5B语言骨干网络,仅需约53M可训练参数,即可在LIBERO、LIBERO-Plus、CALVIN基准及真实世界操作任务中取得强劲性能,显著降低了3D感知VLA的训练与部署门槛。

技术核心包含空间表征蒸馏与轻量级知识路由两大模块。前者将冻结3D专家的空间理解能力迁移至LoRA适配的视觉分支,通过对齐注意力响应并保留原始视觉通路来维持预训练语义知识;后者利用动作损失梯度 derived 的路由监督,选择性地将与动作相关的视觉Token转发给语言模型。该架构避免了推理阶段对额外3D数据的依赖,实现了参数高效的空间自适应学习,相关代码已在GitHub开源供社区复现与验证。

关键看点 ​

  • 空间表征蒸馏:对齐冻结3D专家与LoRA视觉分支注意力响应,在无外部3D输入下迁移空间理解能力
  • 知识路由机制:基于动作损失梯度生成路由监督,精准筛选动作相关视觉Token传输至语言模型
  • 参数高效训练:基于0.5B语言骨干仅微调约53M参数,在LIBERO及真机操作中验证了低资源3D适应性
  • 开源资产释放:完整代码已在GitHub开源,支持具身智能社区进行无3D传感器VLA算法复现与迭代

工程与产业观察 ​

SPARK-VLA验证了通过蒸馏与路由机制可将3D空间理解内化至轻量化VLA中,摆脱了对推理端深度传感器的强依赖。其53M参数的极低微调成本为端侧具身模型部署提供了可行路径,解决了3D感知与语义保持的工程矛盾,对低成本机械臂操作算法研发具有直接参考价值。

英文摘要 (Abstract) ​

Understanding three-dimensional physical structure has become essential for vision-language-action models (VLAs). However, existing approaches either rely on explicit geometric inputs at inference time or adopt a single visual pathway, potentially compromising the visual-semantic knowledge inherited from pre-trained models. We present SPARK-VLA, Spatial Representation distillation and Knowledge routing for VLAs, a preserve-specialize-route framework for spatially adapting VLAs without requiring external 3D inputs or experts at inference. SPARK-VLA consists of two key components: (1) spatial representation distillation, which transfers 3D understanding from a frozen 3D expert to a LoRA-adapted visual branch by aligning attention responses while preserving the original visual pathway, and (2) a lightweight knowledge router, which selectively forwards action-relevant visual tokens to the language model using routing supervision derived from action-loss gradients. Across LIBERO, LIBERO-Plus, CALVIN, and real-world manipulation, SPARK-VLA achieves strong performance with only a 0.5B language backbone and approximately 53M trainable parameters, providing an effective and parameter-efficient approach to 3D-aware VLA learning. Code is available at https://github.com/park-jungin/spark-vla.

作者:Jungin Park, Chaoran Zhu, Changjae Oh, Kwanghoon Sohn

基于本页技术方向关联的整机参数、核心零部件、实测报告与同赛道选型横评。

外部来源

本页事实对应的公开页面,以原文为准。(点击将离开本站)

  1. arXiv 2611.10443:Spatial Representation Distillation and Knowledge Routing for Vision-Language-Action Models ↗访问 2026-11-11