arXiv cs.AI· Nikolaos Tsagkas, Andreas Sochopoulos, Chris Xiaoxuan Lu, Oisin Mac Aodha, Alexandros Kouris·· 6 天前AI 评分29
w²VLA：在视觉-语言-动作模型中解耦陈述性知识与程序性知识
Decoupling the Declarative from the Procedural in Vision-Language-Action Models
AI 导读
研究者提出 w²VLA，一种重构信息流的视觉-语言-动作（VLA）模型，通过以组合且可解释的方式用视觉、空间和技能信息调制机器人状态序列，而非将所有多模态 token 送入不透明的 Transformer 动作专家。该方法成功解耦知识表示，实现稳健的行为克隆，并在差异较大、未见过的物体上取得零样本技能迁移能力。
正文 · 原文
Abstract:Deploying generalist robotic agents in the real world requires transferable skills. Specifically, a policy trained to clone a behavior from object-specific demonstrations must generalize beyond that object, otherwise data collection requirements become intractable. Recently, fine-tuning of pre-trained billion-parameter Vision-Language Models (VLMs), initially on large-scale robot datasets and then on fewer scenario-specific demonstrations, has emerged as the predominant paradigm for designing Vision-Language-Action (VLA) models. While these policies achieve state-of-the-art manipulation performance in-distribution, they remain brittle to minor spatial, semantic, and task variations. In this work, we address the inability of current models to decouple the declarative (i.e., concepts and entity semantics) from the procedural knowledge (i.e., how to do something) encoded in their parameters, which is a fundamental bottleneck for zero-shot skill transfer to novel objects. To address this, we propose w$^{2}$VLA, a new VLA model with restructured information flow. Rather than feeding all multimodal tokens from the VLM encoder into a large, opaque transformer-based action expert, our approach modulates the robot state sequence with visual, spatial, and skill information in a compositional and interpretable manner. Unlike popular, state-of-the-art VLAs, we show that our modular approach successfully decouples knowledge representations, enabling robust behavior cloning and unprecedented zero-shot skill transfer capabilities across dissimilar, unseen objects.
| Subjects: | Robotics (cs.RO); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) |
| Cite as: | arXiv:2606.21496 [cs.RO] |
| (or arXiv:2606.21496v2 [cs.RO] for this version) | |
| https://doi.org/10.48550/arXiv.2606.21496 arXiv-issued DOI via DataCite |
Submission history
From: Nikolaos Tsagkas [view email]
[v1]
Fri, 19 Jun 2026 14:43:58 UTC (25,373 KB)
[v2]
Fri, 9 Oct 2026 12:35:44 UTC (5,502 KB)
来源：arXiv cs.AI · arxiv.org