Qwen-VLA

Qwen-VLA: The official repository of Qwen-VLA 本周 GitHub 最热门开源项目深度解析 项目地址: QwenLM/Qwen-VLA 生成时间: 2026-06-02 20:32:01 项目概览 QwenLM/Qwen-VLA 是本周 GitHub 上最受关注的开源项目之一，在短时间内积累了大量关注。 基本信息 指标 数据 作者 @QwenLM 编程语言 未知 Star 数 413 ⭐ Fork 数 17 创建时间 2026-05-28 最后更新 2026-06-02 项目简介 The official repository of Qwen-VLA 该项目采用多种技术栈构建，具有良好的跨平台兼容性。 核心特性 根据项目 README 分析，Qwen-VLA 的主要特点包括： 高关注度：413 个 Star，说明开发者社区对此项目高度认可 活跃开发：17 个 Fork，社区参与度高 快速成长：自 2026-05-28 创建以来持续获得关注 开源免费：完全开源，可自由使用和二次开发 技术架构 Qwen-VLA 基于 未知 技术栈构建： 编程语言：未知 项目规模：413 个 Star，获得广泛认可 社区活跃度：17 个 Fork，开发者积极参与 README 原文摘要 <div align="center"> <img src="assets/qwen-logo.png" alt="Qwen-VLA" width="260"/> <h1 style="border: none;">Qwen-VLA</h1> <p><b>Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments</b></p> <p align="center"> <b>Qwen Team</b> </p> <p align="center"> <a href="https://arxiv.org/abs/2605.30280">📑 Technical Report</a> | <a href="https://qwen.ai/blog?id=qwenvla">📖 Blog</a> | <a href="https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen-VLA/demo.mp4">🖥️ Demo</a> </p> </div> Welcome to the official repository of **Qwen-VLA**. Here, you can find official information about Qwen-VLA and post your questions ([Issues](https://github.com/QwenLM/Qwen-VLA/issues)). ## 🎬 Demo <div align="center"> <video src="https://github.com/user-attachments/assets/7521d371-a1d5-4743-928d-aa3b5ce7374e" width="100%" controls></video> </div> ## 💡 Introduction <div align="center"> <img src="assets/qwenvla_overview.png" alt="Qwen-VLA Overview" width="90%"/> </div> <br> **Qwen-VLA** is a unified vision-language-action generalist model built upon **Qwen3.5-4B** (vision-language backbone) and a **1.15B DiT flow-matching action decoder**. It casts manipulation, navigation, and trajectory prediction into a shared action-and-trajectory prediction framework, enabling a unified model to learn from heterogeneous embodied data across tasks, environments, and robot embodiments via embodiment-aware prompt conditioning, no per-platform output heads needed. A unified Qwen-VLA generalist **matches or outperforms task-specific specialists** fine-tuned independently per benchmark across multiple simulation and real-world evaluations, pushing embodied intelligence from "skill specialists" toward "generalist actors." ### ✨ Key Highlights - **🏆 One Generalist Beats Specialists.** A unified model matches or outperforms per-benchmark specialists across multiple simulation and real-world evaluations. - **🔗 Unified Action-and-Trajectory Framework.** 最近更新记录 适用场景 Qwen-VLA 适合以下用户： ...