#世界模型 (70)

τ0-VLA: Robot VLA that searches over subtask sequences at test time with a world model

VLA世界模型World Model测试时搜索机器人操作

Galaxea G0.5: One Autoregressive VLA for Reasoning and Action

VLAGalaxeaG0.5视觉语言动作模型自回归

Flex-π: Multi-Stream World-Action Model

VLA世界模型WAM双臂Flex-π

Roblox launches its own world model

世界模型RobloxControlNet预测模型具身AI

Dyna-2 World-Action Model by Dyna Robotics

世界模型世界-动作模型WAMDyna-2Dyna Robotics

DeWorldSG: depth-aware 3D scene graph generation

3D场景图scene graphDeWorldSGECCV 2026RGB-D

NVIDIA ArtiFixer Repairs Incomplete Gaussian Splats

3DGS高斯泼溅空间智能世界模型NVIDIA

FLUX 3: Black Forest Labs launches real world model

FLUX 3Black Forest Labs世界模型world model多模态

Flex-π: compute-flexible robot foundation model

机器人基础模型VLA世界模型3D感知语义感知

Zero-WAM: Show Robots a Video Instead of Explaining New Tasks

世界模型视频示教上下文提示动作生成免微调

Build Manipulation Policies with NVIDIA Cosmos 3 Edge

世界模型world model具身智能NVIDIACosmos

ViTacWorld: Tactile-Enhanced Robot World Model

触觉感知世界模型接触建模ViTacWorldTactile Sensing

DéjàView: Looping Transformers for Multi-View 3D Reconstruction

世界模型多视角重建3D重建NVIDIADVLT

DeWorldSG: Probabilistic 3D Gaussian Scene Graph Generation (ECCV 2026)

3D场景图Scene Graph3DGS世界模型World Model

NVIDIA SimFoundry turns real scenes into simulation worlds

SimFoundryNVIDIA仿真Sim-to-Real世界模型

NVIDIA Cosmos 3 Edge: 4B Edge World Model for Robots

NVIDIACosmos 3 Edge世界模型边缘计算实时机器人

ABot-Recon turns driving video into 3D scenes

ABot-Recon3D重建3D reconstruction空间AIspatial AI

Daimon-TWM: tactile-grounded world model for dexterity

触觉世界模型灵巧操作DaimonWorld Model

ABot World 0.5B World Model on Consumer GPU

世界模型world modelABot World消费级GPU实时

Being-H0.8: Latent Tactile World Action Model

触觉感知世界模型具身智能基础模型Tactile Sensing

World Action Models (WAMs): Unifying VLA and World Models

具身智能Embodied AIVLA世界模型World Models

DeWorldSG: Depth-Aware 3D Semantic Scene Graph Generation via World-Model Priors

3D场景图scene graphRGB-D世界模型V-JEPA 2

Orbis 2: A Hierarchical World Model for Driving

世界模型world model驾驶driving分层

Agentic Real2Sim: Physics-based World Modeling with Vision-Language Agents

世界模型视觉语言模型sim-to-real物理仿真Paper

Koopman Dreamer: Spectrally Constrained Latent Dynamics for Stable World-Model Imagination

世界模型DreamerKoopman谱约束强化学习

GigaBrain-0.7: Scaling Embodied Foundation Models to Emergent Capabilities with a Three-System Architecture

VLA具身智能世界模型GigaBrainFlow Matching

Do Coding Agents Need Executable World Models, Simplification, and Verification to Solve ARC-AGI-3?

ARC-AGI编程智能体coding agent世界模型world model

AeroAct: Action-Centered World-Action Models for Language-Conditioned Quadrotor Flight

无人机UAV四旋翼quadrotor世界模型

Streaming Multi-Agent Autoregressive Diffusion Model with World State Registers

世界模型多智能体扩散模型自回归Paper

PhysCoRe: Physics-Corrected Residual World Models for Material-Aware Deformable Dynamics

世界模型物理仿真可变形物体机器人操作Paper

GS-Agent: Creating 4D Physical Worlds With Generative Simulation

4D生成世界模型3DGS物理仿真生成式

Robot-Factored World Models via Robot Rendering

世界模型World Model机器人渲染Robot Rendering视频

False Prophets: On the Security of World Models in Agentic Systems

Paper世界模型World Model

Real-Time Human-Centric World Modeling for Upper-Body Human-Object Interaction

Paper世界模型World Model

Persistent Computational State: A Session-Centric Runtime for Generative World Models

世界模型World Model生成式Generative运行时

On the Identifiability of Controlled World Models

世界模型World Model可辨识性Identifiability受控

The Physics of Multi-Turn Long-Horizon Planning: From Pre-training to Post-training via Single- and Multi-Teacher On-Policy Agentic Distillation

Paper世界模型World Model

HyWorldVLA: A Vision-Language-Action Model with Hybrid World Modeling for Autonomous Driving

VLA世界模型自动驾驶混合建模Paper

PAVXploreRL: Physical-Action-Visual World Model Reinforcement Learning with Action Exploration

世界模型强化学习动作探索具身智能Paper

PhyAI: Real-Time Physical AI at the Edge, Scalable Rollouts in the Cloud

Physical AI推理引擎VLA世界模型边缘部署

PhyLatent: Learning Dynamics-Relevant Representations for JEPA World Models

JEPAWorld ModelsMPC机器人操控世界模型

Robust-WAM: Bridging Generative Pretraining and Semantic Foresight in World-Action Models

世界模型世界-动作模型视频生成VAE语义前瞻

ω-0: A Latent Predictive World Action Model for Concurrent Humanoid Loco-Manipulation

人形机器人世界模型移动操作全身控制扩散策略

DSWorld: A Data Science World Model for Efficient Autonomous Agents

世界模型world model数据科学data science自主智能体

StageWAM: Joint-Embedding Stage Prediction for World-Action Models in Robot Manipulation

机器人操作世界模型JEPAWAMVLA

DECOWAM: Decoupled Whole-Body World-Action Model for Legged Mobile Manipulation

世界模型VLA移动操作四足机器人全身控制

TRW: TRACE-RealWorld---An Auditable Consistency Contract for World Models as Materialized Views

世界模型World Model一致性Consistency可审计

Zero-WAM: In-Context World-Action Modeling from Human Videos for Open-Ended Task Generalization

世界模型上下文学习人类视频示教操作泛化视频-动作模型

LeVJEPA: Efficient & Scalable Video Pretraining without the Heuristics

视频自监督预训练JEPA表征坍缩SIGReg块因果注意力

Towards Dual-Brain Minimal Sufficient Representation for Vision-Language Navigation

Paper世界模型World Model

WorldDiT: A Unified Diffusion Architecture for World and Action Modeling

Paper世界模型World Model

The World Model Remembers, the Actor Forgets: Dream Rehearsal for Continual Model-Based RL

世界模型持续学习强化学习Dreamer灾难性遗忘

FeelWorld: Visuo-Tactile World Model for Hierarchical Contact Prediction and Planning

Paper世界模型World Model

ViTacWorld: Scaling Visuo-Tactile World Models for Contact-Rich Robot Manipulation

世界模型World Model视触觉Visuo-Tactile机器人操作

World Translation: Minimizing Sim-to-Real Gap with Backward Dynamics Extraction and Unpaired Domain Translation

sim-to-real域适应世界模型反向动力学Paper

GigaBrain-WBC-0.5: A Behavior World Model for Robust Whole-Body Control with Environment Interaction

人形机器人Humanoid世界模型World Model全身控制

Music-JEPA: Learning a World Model of Sound from Action

世界模型World Model声音SoundJEPA

Dreamer-CPC: Message Learning with World Models for Decentralized Multi-agent Reinforcement Learning

世界模型多智能体强化学习Dreamer通信

Action from Adjacent Set in Physical Space Outperforms the Best Prediction in World Models

Paper世界模型World Model

Riemann-1.0: An Embodied World Action Model for Physical AI

具身智能World Action Model因果自回归渐进式预训练Riemann Dynamics

Perceptron Open-Sources Isaac 0.5: First Open Model at the Frontier of Video Understanding, Embodied Reasoning and Robot Control — 210x Less Teleop via Video Scaling

具身基础模型embodied foundation model扩展定律scaling lawVLA

GRASP: Gradient-Based Planning for World Models

世界模型规划梯度优化长程规划World Model

1X World Model: Simulating the Future to Evaluate Robot Policies

世界模型1XNEO机器人评估action-controllable

1X World Model: From Video to Action — A New Way Robots Learn

世界模型1XNEOvideo-to-action具身智能

Dyna-2: A 1-Million-Hour Scaling Law for World-Action Models

世界模型world-action model扩展定律scaling law具身智能

#世界模型 (70)

Google Genie 3 Simulates the Real World from Street View Images

EchoWM: A Drivable World Model, Not Just a Clip

LingBot-VA: Causal Video Model for Robot Control

Wayve Releases GAIA-4 World Model for Robotics Simulation

RoFacto: Action-Conditioned Diffusion World Model

τ0-VLA: Robot VLA that searches over subtask sequences at test time with a world model

Galaxea G0.5: One Autoregressive VLA for Reasoning and Action

Flex-π: Multi-Stream World-Action Model

Roblox launches its own world model

Dyna-2 World-Action Model by Dyna Robotics

DeWorldSG: depth-aware 3D scene graph generation

NVIDIA ArtiFixer Repairs Incomplete Gaussian Splats

FLUX 3: Black Forest Labs launches real world model

Flex-π: compute-flexible robot foundation model

Zero-WAM: Show Robots a Video Instead of Explaining New Tasks

Build Manipulation Policies with NVIDIA Cosmos 3 Edge

ViTacWorld: Tactile-Enhanced Robot World Model

DéjàView: Looping Transformers for Multi-View 3D Reconstruction

DeWorldSG: Probabilistic 3D Gaussian Scene Graph Generation (ECCV 2026)

NVIDIA SimFoundry turns real scenes into simulation worlds

NVIDIA Cosmos 3 Edge: 4B Edge World Model for Robots

ABot-Recon turns driving video into 3D scenes

Daimon-TWM: tactile-grounded world model for dexterity

ABot World 0.5B World Model on Consumer GPU

Being-H0.8: Latent Tactile World Action Model

World Action Models (WAMs): Unifying VLA and World Models

DeWorldSG: Depth-Aware 3D Semantic Scene Graph Generation via World-Model Priors

Orbis 2: A Hierarchical World Model for Driving

Agentic Real2Sim: Physics-based World Modeling with Vision-Language Agents

Koopman Dreamer: Spectrally Constrained Latent Dynamics for Stable World-Model Imagination

GigaBrain-0.7: Scaling Embodied Foundation Models to Emergent Capabilities with a Three-System Architecture

Do Coding Agents Need Executable World Models, Simplification, and Verification to Solve ARC-AGI-3?

AeroAct: Action-Centered World-Action Models for Language-Conditioned Quadrotor Flight

Streaming Multi-Agent Autoregressive Diffusion Model with World State Registers

PhysCoRe: Physics-Corrected Residual World Models for Material-Aware Deformable Dynamics

GS-Agent: Creating 4D Physical Worlds With Generative Simulation

Robot-Factored World Models via Robot Rendering

False Prophets: On the Security of World Models in Agentic Systems

Real-Time Human-Centric World Modeling for Upper-Body Human-Object Interaction

Persistent Computational State: A Session-Centric Runtime for Generative World Models

On the Identifiability of Controlled World Models

The Physics of Multi-Turn Long-Horizon Planning: From Pre-training to Post-training via Single- and Multi-Teacher On-Policy Agentic Distillation

HyWorldVLA: A Vision-Language-Action Model with Hybrid World Modeling for Autonomous Driving

PAVXploreRL: Physical-Action-Visual World Model Reinforcement Learning with Action Exploration

PhyAI: Real-Time Physical AI at the Edge, Scalable Rollouts in the Cloud

PhyLatent: Learning Dynamics-Relevant Representations for JEPA World Models

Robust-WAM: Bridging Generative Pretraining and Semantic Foresight in World-Action Models

ω-0: A Latent Predictive World Action Model for Concurrent Humanoid Loco-Manipulation

DSWorld: A Data Science World Model for Efficient Autonomous Agents

StageWAM: Joint-Embedding Stage Prediction for World-Action Models in Robot Manipulation

DECOWAM: Decoupled Whole-Body World-Action Model for Legged Mobile Manipulation

TRW: TRACE-RealWorld---An Auditable Consistency Contract for World Models as Materialized Views

Zero-WAM: In-Context World-Action Modeling from Human Videos for Open-Ended Task Generalization

LeVJEPA: Efficient & Scalable Video Pretraining without the Heuristics

Towards Dual-Brain Minimal Sufficient Representation for Vision-Language Navigation

WorldDiT: A Unified Diffusion Architecture for World and Action Modeling

The World Model Remembers, the Actor Forgets: Dream Rehearsal for Continual Model-Based RL

FeelWorld: Visuo-Tactile World Model for Hierarchical Contact Prediction and Planning

ViTacWorld: Scaling Visuo-Tactile World Models for Contact-Rich Robot Manipulation

World Translation: Minimizing Sim-to-Real Gap with Backward Dynamics Extraction and Unpaired Domain Translation

GigaBrain-WBC-0.5: A Behavior World Model for Robust Whole-Body Control with Environment Interaction

Music-JEPA: Learning a World Model of Sound from Action

Dreamer-CPC: Message Learning with World Models for Decentralized Multi-agent Reinforcement Learning

Action from Adjacent Set in Physical Space Outperforms the Best Prediction in World Models

Riemann-1.0: An Embodied World Action Model for Physical AI

Perceptron Open-Sources Isaac 0.5: First Open Model at the Frontier of Video Understanding, Embodied Reasoning and Robot Control — 210x Less Teleop via Video Scaling

GRASP: Gradient-Based Planning for World Models

1X World Model: Simulating the Future to Evaluate Robot Policies

1X World Model: From Video to Action — A New Way Robots Learn

Dyna-2: A 1-Million-Hour Scaling Law for World-Action Models