具身开发者社区

外观

论文arXiv 2610.1474618 天前

BridgeWAM：通过 Latent Bridge Queries 桥接视频与动作专家 ​

BridgeWAM用32个LBQ在共享潜在空间桥接多模态先验与动作专家，并在LIBERO、LIBERO-Plus及真机延迟上验证执行、泛化与效率。

AI 整理
  • 感知与导航
  • 仿真与训练
  • 具身模型

原文链接：BridgeWAM: Bridging Video and Action Experts via Latent Bridge Queries（arXiv 2610.14746 · 发布日期：2026-10-10）

BridgeWAM: Bridging Video and Action Experts via Latent Bridge Queries

BridgeWAM用32个LBQ在共享潜在空间桥接多模态先验与动作专家，并在LIBERO、LIBERO-Plus及真机延迟上验证执行、泛化与效率。

核心要点 ​

  • 现有WAM以密集逐层MoT注意力耦合视频与动作专家，易引入无关视觉特征和冗余计算。
  • BoE/BridgeWAM用32个LBQ聚合视觉、语言和本体先验，动作专家通过交叉注意力选择相关信息。
  • 5.06B模型在LIBERO达98.25%成功率，在LIBERO-Plus合并成功率为72.24%。
  • 相较Fast-WAM，单样本推理FLOPs降28.72%，真机延迟77.1 ms、提速2.47倍。

验证范围与局限：验证范围包括LIBERO与LIBERO-Plus任务、仿真FLOPs分析、真机延迟测试及扰动下的定性实验。

英文摘要 (Abstract) ​

World Action Models (WAMs) commonly couple video and action experts through dense, layer-wise Mixture-of-Transformers (MoT) attention, exposing action prediction to irrelevant visual features while enforcing matched expert depths. This coupling hinders selective visual-feature use and introduces redundant computation. We introduce Bridge-of-Experts (BoE), instantiated as BridgeWAM, which aggregates visual, language, and proprioceptive priors through 32 Latent Bridge Queries (LBQs) into a shared, limited latent space. Joint video-action supervision lets action gradients guide the Video Expert toward representations better suited for action prediction, while the Action Expert learns to selectively retrieve task-relevant information through cross-attention. Decoupling expert depths and reusing LBQs during action denoising yield a 5.06B model with a 30-layer Video Expert and a two-layer Action Expert, achieving 98.25% success on LIBERO and 72.24% pooled success on LIBERO-Plus, demonstrating effective task execution and strong generalization with an ultralight action head. Ablations support joint representation adaptation and unified LBQ conditioning. Relative to Fast-WAM, BridgeWAM reduces single-sample inference FLOPs by 28.72% in simulation profiling and real-robot latency to 77.1 ms (a 2.47-fold speedup). Qualitative experimental results further indicate preferential attention to target objects and task destinations across perturbations.

作者：Jiaxin Fang, Ruitong Liu, Zeyu Lai, Xingyu Wang, Mingyuan Jia, Pingyue Sheng, Weikang Wan, Mengdi Xu, Yuxiang Gao

来源

本页事实对应的公开页面，以原文为准。

  1. arXiv 2610.14746：BridgeWAM: Bridging Video and Action Experts via Latent Bridge Queries访问 2026-10-21