外观
论文arXiv 2610.1228520 天前
PLaW-VLA:预测式潜在世界模型提升长程控制与推理效率
本文由 AI 根据公开内容整理,可能有误,请以原文为准:PLaW-VLA: Predictive Latent World Modeling for Vision-Language-Action Policies(arXiv 2610.12285)
核心内容
针对视觉-语言-动作(VLA)策略在长程控制中缺乏预测上下文的问题,研究提出PLaW-VLA方法。该方法摒弃对控制无关视觉细节的重建,转而在预训练的预测导向表征空间中建模任务相关的未来状态。实验数据显示,该模型在RoboTwin Hard Horizon III基准上较反应式策略提升11.8个百分点,在零样本LIBERO-Plus测试中较重构导向潜在预测提升1.77个百分点,显著增强了分布偏移下的泛化能力。
技术实现上,PLaW-VLA采用Mixture-of-Transformers架构,利用结构化因果注意力机制将观测历史、当前任务语义及预测的未来状态作为动作生成的条件输入。通过避免低层级视觉重建,该模型实现了并行未来预测,在保持同等策略性能的前提下,其推理延迟仅为生成式世界-动作建模方法的约1/19,大幅降低了具身智能体端侧部署的计算负载,为实时长程任务规划提供了高效解决方案。
关键看点
- 长程控制增益:在RoboTwin Hard Horizon III基准测试中,PLaW-VLA相比传统反应式策略任务成功率提升11.8个百分点,有效解决长时序依赖问题。
- 推理效率突破:通过潜在空间预测替代像素级重建,实现并行计算,推理延迟降至生成式世界模型方法的约1/19,满足实时控制需求。
- 架构设计创新:基于Mixture-of-Transformers与结构化因果注意力,融合观测历史、任务语义及预测状态,强化动作生成的上下文条件约束。
- 零样本泛化验证:在LIBERO-Plus零样本测试中较重构导向方法提升1.77pp,证实预测任务相关表征比视觉重建更能适应分布偏移场景。
工程与产业观察
PLaW-VLA证明了在VLA策略中引入“预测导向”而非“重建导向”的潜在世界模型是平衡性能与算力的关键路径。对于具身智能工程落地而言,这种将世界模型压缩至潜在空间并解耦视觉生成的范式,能显著降低端侧推理门槛,为解决长程操作任务中的实时性与泛化性矛盾提供了可复用的算法参考。
英文摘要 (Abstract)
Learning to predict how the world evolves can provide vision-language-action (VLA) policies with predictive context for long-horizon control, but its effectiveness depends on what future representation is modeled and how it conditions action generation. We introduce PLaW-VLA, which models task-relevant future states in a pretrained prediction-oriented representation space, reducing the need to predict control-irrelevant visual details. Built on a Mixture-of-Transformers architecture, PLaW-VLA conditions action generation on observation history, current task semantics, and predicted future states through structured causal attention. Experiments show a +11.8 percentage-point (pp) gain over reactive policies on RoboTwin Hard Horizon III and a +1.77 pp gain over reconstruction-oriented latent prediction on zero-shot LIBERO-Plus, supporting improved long-horizon control and generalization under distribution shift, respectively. By avoiding low-level visual reconstruction, PLaW-VLA lowers the burden of future prediction, enabling a lightweight latent world model with parallel future prediction and about 1/19 the inference latency of generative world-action modeling at comparable policy performance.
作者:Yu Liu, Hetian Guo, Tianlv Huang, Ziyi Cai, Wudi Chen, Hantang Wang, Qiutong Liu, Yingzhi Peng, Wei Han, Peijun Tang, Jianan Wang, Zipei Fan, Zhiyuan Zha, Xuan Song
来源
本页事实对应的公开页面,以原文为准。