1
2
华为具身大脑一号位做类脑智能世界模型，对标JEPA，获亿元级融资｜硬氪首发
华为具身大脑关键人物朱森华以类脑智能范式打造认知世界模型，获亿元融资，对标Yann LeCun的JEPA路径。
作者｜黄楠 编辑｜袁斯来 硬氪获悉，具身智能大脑公司「具脑磐石」完成新一轮亿元级融资，本轮融资由具备深厚类脑与具身产业背景的顶尖产业资本领投，老股东及多家顶尖基金复投和跟投，多维资本担任独家财务顾问。同时，更新一轮融资也在同步交割中。 资金将重点投入核心技术研发、人才团队扩容及全球化市场拓展，以加速…
3
4
“VLA和世界模型都不是终局，会有物理世界独有的模型” | 蚂蚁灵波沈宇军@AIGC2026
蚂蚁灵波沈宇军提出AI 2.0下半场应从内容生成转向动作生成，认为VLA和世界模型都不是终局，物理世界需要独特模型。
“要做机器人时代的安卓系统”
5
Roundtables: Can AI Learn to Understand the World?
AI能否真正理解世界？MIT权威圆桌聚焦2026年世界模型新突破
Listen to the session or watch below AI companies want to build systems that understand the external world and overcome the limitations of LLMs. Recen…
6
城市级AI服务：从试点到常态化，机器人的实景作战与规模化落地| 2026AI Partner·北京亦庄AI+产业大会
酷哇以“以战养战”模式，驱动机器人真实运营中迭代，已在50+城市实现环卫、出行多场景规模化落地，具身智能加速破圈。
当Robotaxi还在为L4苦苦挣扎时，酷哇的环卫机器人、无人小巴、机器狗已经在50多个城市“上岗”赚钱了。 具身智能最大的瓶颈不是算法，而是数据——没有量产就没有数据，没有数据就无法进化。酷哇的解法是“以战养战”：让机器人在真实运营中一边干活一边成长，用万台规模反哺模型迭代。李柯宏强调，中国是全球…
7
PhyWorld: Physics-Faithful World Model for Video Generation
为视频生成注入物理规律，让模型理解重力、碰撞等真实性，推动世界模型迈向新高度
arXiv:2605.19242v1 Announce Type: cross Abstract: World simulators can provide safe and scalable environments for training Physical AI systems before …
8
OrbiSim: World Models as Differentiable Physics Engines for Embodied Intelligence
将世界模型重构为全可微物理引擎，为具身智能提供统一物理约束路径。
arXiv:2605.16395v1 Announce Type: cross Abstract: We present OrbiSim, a novel robotic simulation paradigm that redefines world models as a fully diffe…
9
ARROW: Augmented Replay for RObust World models
提出ARROW增强回放框架，显著提升世界模型在分布外场景的鲁棒性。
arXiv:2603.11395v2 Announce Type: replace Abstract: Continual reinforcement learning challenges agents to acquire new skills while retaining previousl…
10
PH-Dreamer: A Physics-Driven World Model via Port-Hamiltonian Generative Dynamics
物理驱动的世界模型新范式，Port-Hamiltonian 动力学让生成式 AI 更贴近真实物理规律。
arXiv:2605.18303v1 Announce Type: new Abstract: World models built on recurrent state space architectures enable efficient latent imagination, yet rem…
11
Stera: Open-Source Infra That Turns iPhones into Spatial Data for World Models
开源项目Stera将普通iPhone升级为研究级空间数据采集系统，并开源10M帧数据集，为具身AI世界模型提供高质量训练数据。
We are releasing Project Stera - an open source, end-to-end pipeline that turns a commodity iPhone into a research-grade capture system for embodied A…
12
Quantitative Video World Model Evaluation for Geometric-Consistency
提出PDI-Bench框架，量化评估生成视频模型的几何一致性，攻克3D结构合理性难题。
arXiv:2605.15185v1 Announce Type: cross Abstract: Generative video models are increasingly studied as implicit world models, yet evaluating whether th…
13
Agentifying Patient Dynamics within LLMs through Interacting with Clinical World Model
基于LLM与临床世界模型交互，模拟患者动态以实现ICU脓毒症序贯治疗决策。
arXiv:2605.14723v1 Announce Type: new Abstract: Sepsis management in the ICU requires sequential treatment decisions under rapidly evolving patient ph…
14
RoboWM-Bench: A Benchmark for Evaluating World Models in Robotic Manipulation
首个专为机器人操作评估世界模型物理一致性的基准，填补视觉真实与可执行性之间的鸿沟。
arXiv:2604.19092v2 Announce Type: replace-cross Abstract: Recent advances in large-scale video world models have enabled increasingly realistic future…
15
Identifiable Token Correspondence for World Models
提出一种可识别token对应方法用于世界模型，强化智能体对环境结构的泛化理解
arXiv:2605.16457v1 Announce Type: new Abstract: Transformer-based world models have shown strong performance in visual reinforcement learning, but oft…
16
Google’s Genie world model can now simulate real streets with Street View
Google基于20年2800亿张街景图像训练Genie世界模型，能模拟真实街道场景，开启沉浸式AI世界生成。
Google DeepMind is integrating Street View with Project Genie to create immersive, interactive world simulations for robotics, gaming, and travel, all…
17
Composition of Memory Experts for Diffusion World Models
快速访问arXiv上这篇融合记忆专家与扩散世界模型的前沿论文，一键获取AI研究最新突破与全文PDF。
arXiv:2605.18813v1 Announce Type: new Abstract: World models aim to predict plausible futures consistent with past observations, a capability central …
18
Agora-1 by Odyssey
多智能体世界模型，既能玩又能讨论，开启沉浸式互动新体验。
A multi-agent world model you can play Discussion | Link
19
DiLA: Disentangled Latent Action World Models
DiLA将潜在动作解耦为几何与纹理流，实现高保真视频预测，突破LAMs的抽象-保真权衡。
arXiv:2605.15725v1 Announce Type: cross Abstract: Latent Action Models (LAMs) enable the learning of world models from unlabeled video by inferring ab…
20
Latent Video Prediction Learns Better World Models
首次系统评估自监督视频模型作为世界模型，揭示潜在视频预测能学习更优的世界模型。
arXiv:2605.15618v1 Announce Type: cross Abstract: Self-supervised video models are increasingly framed as world models, yet their evaluation remains l…