PulseAugur
实时 09:04:41

Being-H0.7 model integrates future reasoning into robot control without visual rollouts

Researchers have introduced Being-H0.7, a novel latent world-action model designed to enhance robot control by integrating future prediction without generating explicit future video frames. This model utilizes learnable latent queries as a reasoning interface, trained using a dual-branch approach that aligns current context embeddings with those derived from future observations. By focusing on latent space alignment, Being-H0.7 enables policies to reason about future states and actions efficiently, achieving state-of-the-art performance across various simulation and real-world robotic tasks. AI

影响 Introduces a more efficient method for robots to predict future states and actions, potentially improving real-world task performance.

排序理由 This is a research paper detailing a new model for robot control.

在 arXiv cs.CV 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。 我们如何撰写摘要 →

Being-H0.7 model integrates future reasoning into robot control without visual rollouts

报道来源 [1]

  1. arXiv cs.CV TIER_1 English(EN) · Hao Luo, Wanpeng Zhang, Yicheng Feng, Sipeng Zheng, Haiweng Xu, Chaoyi Xu, Ziheng Xi, Yuhui Fu, Zongqing Lu ·

    Being-H0.7: A Latent World-Action Model from Egocentric Videos

    arXiv:2605.00078v1 Announce Type: cross Abstract: Visual-Language-Action models (VLAs) have advanced generalist robot control by mapping multimodal observations and language instructions directly to actions, but sparse action supervision often encourages shortcut mappings rather …