从基础到应用：在实践中改进VLA模型
From Foundation to Application: Improving VLA Models in Practice
AI总结：
针对VLA模型实验室与实际应用差距问题，提出LingBot-VLA 2.0。通过改进数据处理、扩展动作空间及预测动力学建模等方法，经GM-100基准测试验证，提升了模型跨实体长距离移动操作能力。
摘要：
尽管VLA基础模型最近取得了进展，但实验室条件与实际应用之间的差距仍阻碍其实际实施。为弥合这一差距，我们提出了LingBot-VLA 2.0，它在三个功能领域进行改进。一是跨任务和实体的泛化，改进数据处理管道并精心策划约60000小时预训练数据；二是除双臂硬件平台外扩展动作空间；三是用于改进时间推理的预测动力学建模。在GM-100基准测试中的评估验证了这些修改的有益影响，LingBot-VLA-2.0展示了强大的跨实体长距离移动操作能力。
英文摘要：
Despite recent progress of VLA foundation models, the disparity between laboratory conditions and real-world applications continues to impede their practical implementation. To bridge this gap, we present LingBot-VLA 2.0, which advances LingBot-VLA through improvements in three functional domains. (1) Generalization across tasks and embodiments. Compared to the previous version, we revamp the data processing pipeline and curate around 60,000 hours of data for pretraining, including 50,000 hours of robot trajectories spanning 20 robot configurations and 10,000 hours of egocentric human videos. (2) Expanded action space in addition to dual-arm hardware platforms. In particular, our system accommodates degrees of freedom for the heads, waists, mobile bases, and dexterous hands, thereby empowering the robots to tackle more complex tasks in practical scenarios. (3) Predictive dynamics modeling for improved temporal reasoning. Specifically, we formulate future prediction as a proxy task, facilitated by a video representation model for semantic priors and a depth estimation model for geometric cues. Evaluations on the GM-100 benchmark, conducted in a generalist setting, validate the beneficial impact of these proposed modifications. Furthermore, benefiting from the expanded pretraining data that covers whole-body degrees of freedom, LingBot-VLA-2.0 demonstrates strong cross-embodiment long-horizon mobile manipulation capability across the two robotic platforms.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
