外观
论文HF Daily Papers · arXiv 2610.0523024 天前
Vela模型采用自适应样条参数化突破VLA动作表征瓶颈
本文由 AI 根据公开内容整理，可能有误，请以原文为准：Vela: Scaling Vision-Language-Action Models with Adaptive Action Curve Parametrization（HF Daily Papers · arXiv 2610.05230）
核心内容
现有视觉-语言-动作（VLA）模型多采用固定速率动作块表征未来运动，导致输出预算受限，难以兼顾长时域覆盖与接触式操作所需的局部精度。Vela模型通过引入连续轨迹表征替代离散点预测，解决了相邻动作高度相关造成的容量浪费问题，并在LIBERO-X、EBench仿真基准及煎蛋饼、土豆丝两项真实世界长程操作任务中取得优异性能，验证了连续动作表征作为具身基础模型底座的潜力。
技术上，Vela结合紧凑的样条动作表征与运动依赖的时间支撑机制，使固定输出预算能根据运动特性自适应调整时序分辨率，并设计了共享动作接口以适配异构本体。该方法将轨迹平滑性从隐式学习转为显式建模，有效缓解了传统Pointwise表征在长视野与高精度间的权衡困境。研究团队已在项目页开源相关成果，为下一代具身智能模型提供了新的动作空间建模范式。
关键看点
- 连续轨迹表征：采用样条曲线替代固定速率动作块，消除相邻动作冗余，显式保证时序连续性
- 自适应时序分辨率：引入运动依赖时间支撑机制，使固定输出预算随运动快慢动态分配采样密度
- 异构本体兼容：设计共享动作接口统一多形态机器人输出空间，支持大规模多本体数据联合预训练
- 真实任务验证：在煎蛋饼与土豆丝等接触密集长程操作中表现优异，证实连续表征优于离散Chunking
工程与产业观察
Vela证明了将机器人动作空间从离散Token回归连续函数表征的工程价值。对于具身算法工程师，样条参数化不仅降低了高频控制数据的存储与推理开销，更通过结构化归纳偏置提升了接触丰富任务的泛化能力。这提示行业在追求大模型Scaling Law的同时，应重新审视动作空间本身的数学表达效率，而非单纯堆叠Transformer层数。
英文摘要 (Abstract)
Most vision-language-action models represent future motion as fixed-rate action chunks, tying temporal resolution and prediction horizon to a fixed output budget. This pointwise representation wastes capacity on highly correlated neighboring actions, leaves temporal continuity and smoothness to be learned implicitly, and forces a tradeoff between long-horizon coverage and the local precision required for contact-rich manipulation. To address these limitations, we introduce Vela, a vision-language-action foundation model that represents future robot behavior as continuous trajectories. Vela combines a compact spline-based action representation with motion-dependent temporal support and a shared action interface for heterogeneous embodiments, allowing a fixed output budget to adapt its temporal resolution across motions. We pretrain Vela on large-scale multi-embodiment robot data and evaluate it on LIBERO-X, EBench, and two real-world long-horizon tasks, egg-cake cooking and potato shredding, obtaining promising results across simulation and physical manipulation. These results highlight the potential of continuous action representations as a foundation for future embodied foundation models. Project page and more results: https://clementine24.github.io/Vela/ .
Hugging Face Daily Papers 热度：4 upvotes 作者：Yifan Li, Jiaxu Wang, Dongming Wu, Yicheng Jiang, Ryan Ji, Xiangyu Yue, Yanwei Fu
来源
本页事实对应的公开页面，以原文为准。