RAGu 用检索增强让 VLA 机器人成功率从 20% 提升到 80%
Zero-Shot VLA Adaptation with Retrieved Guidance
把 RAG 的思路搬到了机器人上：不重训模型，直接检索 DROID 里类似任务的演示轨迹，几何对齐后用来引导 flow-matching 推理。真实环境成功率从 20% 干到 80%，做机器人操作的工程师值得看看这套免训练适配思路。
arXiv 论文提出 RAGu，一种面向 Vision-Language-Action（VLA）模型的检索增强方法，无需重训基础模型即可完成零样本跨域适配。在真实世界测试中，它将 VLA 的操作成功率从 20% 提升到 80%，并在仿真基准上超过最接近的相关基线方法。
技术上，RAGu 针对机器人领域的检索难题做了三项设计：用几何标定把源域演示轨迹变换到目标环境，缩小跨域差异；采用粗到细的检索策略，从大量预变换候选中选出与基础 VLA 最兼容的轨迹；在推理阶段引导模型的 flow-matching 过程，而不是微调或扩上下文窗口。实验覆盖真实环境、仿真、sim-to-real 迁移和跨本体的人形机器人操作，检索源数据来自 DROID 这类与目标域不匹配的 in-the-wild 演示集。
这项工作把语言模型中的 RAG 思路移植到机器人操作，表明 VLA 面对相机视角、背景、物体和机器人本体变化时，可以通过检索外部演示数据弥补泛化短板，对不想承担重训成本的机器人开发者有直接参考价值。
Zero-Shot VLA Adaptation with Retrieved Guidance
Pre-training large parametric "Vision-Language-Action" (VLA) models on large demonstration datasets has recently yielded promising results for general-purpose manipulation. However, these models still struggle under even small domain shifts relative to the training data, such as changes in camera viewpoint, background, object identity, or robot morphology. Motivated by the success of Retrieval-Augmented Generation (RAG) in language models, we pursue retrieval to boost VLA performance. Specifically, we seek to leverage retrieval from domain-mismatched, "in-the-wild" demonstration datasets like DROID to guide a VLA's behavior in a new target domain. To achieve this, we introduce RAGu, which overcomes robotics-specific challenges for RAG: (1) It incorporates simple geometric calibration to transform source demonstrations into the target setup and reduce cross-domain differences. (2) It uses a coarse-to-fine retrieval strategy that selects, from a large pool of pre-warped candidates, the trajectory most compatible with the base VLA. (3) It guides the VLA's flow-matching inference procedure rather than seeking to retrain the base model or expand its limited context window. We evaluate RAGu across real-world, simulation, sim-to-real transfer, and cross-embodiment humanoid manipulation. We find it improves real-world VLA success rates from 20% to 80% and outperforms the closest-related baselines in sim. Project website: raguvla.github.io
本条情报归属 AITOP「论文」核心栏目，重点聚焦北美前沿实验室与硅谷科技巨头。在跨区域产业观察坐标中，该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线，反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括：VLA、RAG、DROID、机器人操作，读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明：本文原始线索采编自「arXiv cs.LG」（@Subin Kim, Anh-Quan Pham, Dinesh Jayaraman），经 AITOP 多维评分管线结构化提炼。