https://auto-driving-blog.vercel.app/tags/-%E6%A8%A1%E5%9E%8B%E5%AF%B9%E6%AF%94/ Recent content in 📊 模型对比 on Elon's AD Insight https://auto-driving-blog.vercel.app/images/share.png https://auto-driving-blog.vercel.app/images/share.png Hugo zh-cn Sun, 19 Jul 2026 00:00:00 +0000 https://auto-driving-blog.vercel.app/posts/knowledge/%E5%89%8D%E6%B2%BFvla%E6%A8%A1%E5%9E%8B%E5%85%A8%E6%99%AF%E9%80%9F%E8%A7%88/ Sun, 19 Jul 2026 00:00:00 +0000 https://auto-driving-blog.vercel.app/posts/knowledge/%E5%89%8D%E6%B2%BFvla%E6%A8%A1%E5%9E%8B%E5%85%A8%E6%99%AF%E9%80%9F%E8%A7%88/ VLA模型从2023年的RT-2到2026年的GR-3经历了三代演进。本文系统对比15+主流VLA模型的核心设计选择：动作表征（tokenization/连续控制/扩散策略）、视觉编码器架构、LLM backbone选择、以及训练范式（预训练→指令微调→RLHF）。重点分析pi0的Flow Matching动作头、GR-3的CoT推理与自回归架构、BridgeVLA的桥接监督等前沿技术路线。