arxivcs.ROcs.CV2026-07-31
FibVLA: An Efficient Temporal Vision-Language-Action Model with Fibonacci Sampling
Li Lin, Wujun Xu, Weiwei Meng, Kaiwen Xia, Kang Hao Cheong, Shuai Wang
Vision-language-action models (VLAs), which leverage the cognition of multimodal information to infer physical-world actions, provide a generalized solution for embodied AI applications. Conventional VLAs usually concentrate on current digital cognition. While some efforts are ma…