BridgeWAM: Bridging Video and Action Experts via Latent Bridge Queries
Summary
BridgeWAM 通过潜在桥接查询（LBQs）解耦视频与动作专家，提升世界行动模型的动作预测效率与泛化能力，在 LIBERO 上达到 98.25% 成功率，同时将推理 FLOPs 降低 28.72%。
View Cached Full Text
Cached at: 10/19/26, 07:28 AM
Paper page - BridgeWAM: Bridging Video and Action Experts via Latent Bridge Queries
Source: https://huggingface.co/papers/2610.14746
Abstract
BridgeWAM decouples video and action experts via latent bridge queries to improve action prediction efficiency and generalization in world action models.
World Action Models(WAMs) commonly couple video and action experts through dense, layer-wiseMixture-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 introduceBridge-of-Experts(BoE), instantiated as BridgeWAM, which aggregates visual, language, and proprioceptive priors through 32Latent Bridge Queries(LBQs) into a shared, limited latent space.Joint video-action supervisionlets action gradients guide the Video Expert toward representations better suited for action prediction, while the Action Expert learns to selectively retrieve task-relevant information throughcross-attention. Decoupling expert depths and reusing LBQs duringaction denoisingyield a 5.06B model with a 30-layer Video Expert and a two-layer Action Expert, achieving 98.25% success onLIBEROand 72.24% pooled success onLIBERO-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.
View arXiv pageView PDFProject pageGitHub22Add to collection
Get this paper in your agent:
hf papers read 2610\.14746
Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash
Models citing this paper0
No model linking this paper
Cite arxiv.org/abs/2610.14746 in a model README.md to link it from this page.
Datasets citing this paper0
No dataset linking this paper
Cite arxiv.org/abs/2610.14746 in a dataset README.md to link it from this page.
Spaces citing this paper0
No Space linking this paper
Cite arxiv.org/abs/2610.14746 in a Space README.md to link it from this page.
Collections including this paper0
No Collection including this paper
Add this paper to acollectionto link it from this page.
Similar Articles
WAVE-0: Bridging VLA Policies and WAMs with Lightweight Latent Future Prediction
WAVE-0 提出一个轻量级潜在世界模型，将 VLA 策略与世界动作模型（WAM）结合，使预测的动作后果直接引导动作生成，在 RoboTwin2.0 和 LIBERO 系列基准上取得 91.47%/98.7% 的成功率，并将 π0.5 的真实任务成功率从 38.3% 提升至 73.3%。
Light-WAM: Efficient World Action Models with State-Fusion Action Decoding
Light-WAM is a lightweight world action model for efficient robot manipulation that uses a compact video backbone and downsampled latent space for future-video supervision, achieving high performance with low inference latency.
LeWAM: End-to-End World and Action Modeling with JEPAs
LeWAM 提出一种端到端的 JEPA 控制框架，通过联合训练下一隐状态预测与动作流匹配目标，并对隐状态施加各向异性高斯约束，使单一模型同时支持直接策略执行与高效规划，在接触丰富任务和长程任务上将成功率从 28.6% 提升至 89.7% 和从 10.9% 提升至 46.2%，规划速度提升 32.3 倍。
PAIR: Bridging Perception and Action in Vision-Language-Action Models
PAIR 引入显式的「Bridge Tokens」中间表示，将视觉-语言理解与可执行机器人动作连接起来，通过与动作潜变量对齐并提前注入动作专家，在 LIBERO、CALVIN 及真实操作任务上一致提升性能。
Long-WAM: Scaling the Context of World-Action Models
Long-WAM 提出了一种在实时控制约束下扩展因果世界-动作模型上下文的模型-系统框架，发现更长的视觉历史仅在视频基础模型以自回归方式预训练时才显著有效，在 RoboCasa GR-1 上将成功率从 63.3% 提升至 78.7%，并在 Unitree G1 实机上实现 95% 的动态杯子堆叠成功率。