研究人员开发了Zero-WAM，一种新颖的因果视频-动作模型，旨在使机器人能够通过学习人类演示视频来执行未见过的操作任务。这种方法借鉴了大型语言模型的上下文学习，使用视频作为任务规范。为了克服数据稀缺问题，一个自动化流程生成了一个包含74.2K个人类-机器人上下文学习对的数据集。Zero-WAM在模拟中的七个未见过任务上取得了47.0%的成功率，显著优于现有的视频-动作模型，并在现实场景中展示了泛化能力。 AI
影响 通过学习人类视频演示，使机器人能够泛化到新的操作任务，可能加速机器人学习。
排序理由 发布了一篇详细介绍新模型和数据集的新研究论文。
在 Hugging Face Daily Papers 阅读 →
AI 生成摘要 · Google Gemini · 来自 3 个来源。 我们如何撰写摘要 →
完整方法见我们的编辑标准。
Zero-shot cross-task generalization, where a policy must execute manipulation tasks never seen during training, remains a central challenge in robot learning. In large language models, a novel task can be performed simply by specifying it in the context, without any parameter upd…
Zero-WAM enables robotic manipulation of unseen tasks by conditioning a causal video-action model on in-context human video guidance, supported by an automatically generated dataset and a future-chunk prediction objective.
arXiv:2608.26103v1 Announce Type: cross Abstract: Zero-shot cross-task generalization, where a policy must execute manipulation tasks never seen during training, remains a central challenge in robot learning. In large language models, a novel task can be performed simply by speci…