LeCun 的 JEPA 主张、DeepMind Genie 3、NVIDIA Cosmos 与李飞飞的 World Labs，讲清世界模型与视频生成模型的本质区别：一个产出录像，一个产出可交互、可预测的环境。

!(https://www.gstatic.com/images/branding/googleg/1x/googleg_standard_color_128dp.png)Add as a preferredsource on Google…

Tactile sensing is an essential modality for robots performing contact-rich, dexterous manipulation, particularly under …

World modeling enables intelligence to anticipate consequences, guide interventions, and learn from interaction. Yet pre…

This paper shows that JEPA-style latent world models (LWMs) can learn across diverse domains and data sources by adoptin…

LeCun 出走创办 AMI Labs、AMD 并入 World Labs、NVIDIA 开源 Cosmos 3、DeepMind 把 Genie 3 开进订阅渠道——过去十四个月，世界模型从论文词汇变成基础模型新战场。本文按阵营拆解格局、时间线与资本动向，也带上怀疑者的冷水：这可能只是下一个 buzzword。

提出基于JEPA的机器人世界模型RoboJEPA，在覆盖12种机器人本体的数据集上训练，发现其想象误差随算力呈二阶幂律，可预测模型质量与下游规划性能的扩展规律。

该论文提出在无重建、无解码器、无标签的条件下，基于“结构多样性”条件和LeJEPA的线性可识别性，可证明地恢复个体的世界潜变量，弥补JEPA等方法只能识别到线性变换的不足。

We introduce Adaptive LeWorldModel (ALeWM), a world model based on a joint-embedding predictive architecture (JEPA) that…

梦晨* 2026-10-03 10:38:19 来源：量子位 文婷 发自 凹非寺 量子位 | 公众号QbitAI 他来了他来了，TypeSafe AI的联合创始人兼CEO **Diogo Almeida**，顶着一头新染的红发闪亮登场了！…

World-action models give visual navigation agents a way to anticipate how candidate actions will change future observati…

EEG foundation models hold promise for scalable brain-signal decoding across clinical and cognitive neuroscience applica…

《AI 虚拟细胞（AIVC）技术趋势、产业生态与应用前景研究报告——AI 时代生命科学的新型基础设施》正式发布。]

同一预测核心，跨七类系统验证]

3D diffusion policies are strong at generating geometrically grounded actions from current observations, but successful …

World-action models (WAMs) jointly model future observations and actions, typically predicting the future as RGB images.…

Mainstream robotic policies often adopt a Markovian formulation, but many complex real-world manipulation tasks are inhe…

把 2026 年可核查的公开证据整理成一张六层能力路线图——预训练、后训练 RL、推理时计算、上下文与记忆、智能体与 Harness、世界模型。含 Meta ScaleRL 40 万 GPU 小时实验结论、RL 预算占比 10%–30% 口径、Chinchilla 对比、Meta-Harness 6x 差距等数据锚点，并给出优先级表与算法工程师/产品经理的行动建议。