原生全模态技术战略闭环，智象（HiDream.ai）发布具身世界模型HiDream-O1-Embodied
AI 导读
具身世界模型原生全模态扰动适应多视角感知模型数据闭环
智象发布具身世界模型HiDream-O1-Embodied，实现全模态技术闭环，通过强化物理感知与动态预判，登顶具身智能评测榜首，标志向真实世界交互迈入关键阶段。
HiDream.ai launches HiDream-O1-Embodied, achieving full-modal technical closure with enhanced physical perception and dynamic prediction, topping the RoboColiseum benchmark, marking a key step toward real-world interaction.
重点速览
- 模型整合图像/视频/3D/动作等多模态，贯通理解-推演-执行全流程 Integrates image/video/3D/action modalities for end-to-end understanding-execution
- 在扰动适应评测中以0.692分登顶，展现强抗干扰能力 Scores 0.692 in Robustness benchmark, showcasing strong disturbance resistance
- 突破传统指令理解局限，实现意图精准捕捉 Breaks traditional keyword matching for intent recognition
- 多视角协同感知机制提升复杂环境稳定性 Multi-perspective perception enhances complex environment stability
- 创新'模型+数据'策略，构建真实世界训练范式 Innovative model-data synergy builds real-world training paradigm
一句话：HiDream-O1-Embodied通过全模态闭环与抗扰动能力，推动具身智能向真实世界应用突破。 / HiDream-O1-Embodied advances embodied intelligence toward real-world applications through full-modal closure and disturbance resilience.
二次创作声明：本页为 AI 热榜聚合导读，内容与热度数据来自公开来源 （qbitai），版权归原始作者所有；本站仅做转载指引与摘要评述，不复制原文。