| title | BridgeWAM: Bridging Video and Action Experts via Latent Bridge Queries |
|---|---|
| authors | Jiaxin Fang, Ruitong Liu, Zeyu Lai, Xingyu Wang, Mingyuan Jia, Pingyue Sheng, Weikang Wan, Mengdi Xu, Yuxiang Gao |
| date | 2026-10-10 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2610.14746 |
| affiliation | Fudan University |
| upvotes | 44 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| 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. |
cite
@misc{fang2026bridgewam,
title={BridgeWAM: Bridging Video and Action Experts via Latent Bridge Queries},
author={Jiaxin Fang and Ruitong Liu and Zeyu Lai and Xingyu Wang and Mingyuan Jia and Pingyue Sheng and Weikang Wan and Mengdi Xu and Yuxiang Gao},
year={2026},
eprint={2610.14746},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2610.14746}
}