Tag
AffordanceWAM 提出一种感知可供性（affordance）的生成式世界-动作模型，利用人类视频中的可供性表征联合预测未来场景与机器人动作，无需人类动作标注即可实现人类到机器人的迁移，并在 RoboCasa、CALVIN 等基准上取得一致提升。
UNITAS是首个3D原生的世界动作模型（WAM），在共享的度量3D坐标系中统一观测、动作与场景动态，以1.7B参数在RoboTwin场景预测上取得最佳结果（位移误差较PointWorld降低最多49%），并在LIBERO上达到99.8%的成功率。
The paper proposes ViGAR, a hierarchical world-action model for long-horizon compositional robotic manipulation that factorizes tasks into a visual subgoal planner and a subgoal executor, achieving 82.00% and 67.02% success on the RoboTwin Clean2Random benchmark and outperforming the strongest baseline by 12.86 percentage points.
DroneWAM is an efficient world-action model for drone visual navigation that adopts a JEPA-based architecture to model future states directly in representation space, reducing computational costs. It introduces adaptive rollout to allocate prediction depth effectively and achieves best trajectory accuracy on the simulated DroneNav-6D dataset.
InternW0-Δ is a unified world action model that integrates visual dynamics, semantics, geometry, and actions using a Mixture-of-Transformers framework, pretrained on over 20K hours of open data for robot manipulation. It demonstrates strong performance in simulations and real robots and will be open-sourced.
This paper introduces GE-Act 2.0, a world-action model pretrained from scratch to enable scalable zero-shot robotic manipulation with improved success rates across diverse tasks and conditions.
GameWAM introduces the first world-action model for native closed-loop gameplay and GUI control in video games, jointly generating visual observations and executable actions with competitive task success and revealing a source-sensitivity failure mode.
GameWAM introduces the first unified world-action model for native video-game control, jointly predicting future visuals and executable actions using block-causal flow matching and mode-specific distributions.
The paper introduces DECOWAM, a decoupled whole-body world-action model for legged mobile manipulation that improves video and action prediction performance over existing models like FastWAM through dedicated conditional interfaces and a new dataset.
Dyna Robotics introduces Dyna-2, a world-action model pre-trained on one million hours of human video, revealing scaling laws and suggesting the embodiment gap is an adaptation problem rather than a knowledge problem.
Dyna-2 is a world-action model pre-trained on over a million hours of human video, showing scaling laws on human data and a human-to-robot transfer scaling law for zero-shot robot performance.
Dyna Robotics introduces Dyna-2, a world-action model pre-trained on one million hours of human video, discovering new scaling laws for robot manipulation.
SimWAM is a simple yet effective World Action Model for end-to-end autonomous driving that uses video generation purely as a training signal, achieving state-of-the-art 91.5 PDMS on NAVSIM while reducing inference latency.
ST-WAM proposes a semantic-temporal world action model that uses DINOv3 features as a shared semantic representation to improve robot manipulation robustness under visual distribution shifts, achieving 98.7% on LIBERO and 92.8% on RoboTwin 2.0, with significant gains over Fast-WAM in zero-shot settings.
ABot-M0.5 is a new World Action Model for mobile manipulation that improves performance through temporal granularity alignment, action space disentanglement, and train-test consistency, achieving state-of-the-art results on long-horizon and fine-grained manipulation benchmarks.
World Pilot enhances Vision-Language-Action models by incorporating dynamic scene evolution and trajectory priors from a World-Action Model, achieving state-of-the-art zero-shot performance on manipulation tasks.
AHA-WAM is an asynchronous world-action model that uses dual Diffusion Transformers to decouple world prediction from action execution, achieving efficient long-horizon planning and real-time control. It achieves state-of-the-art performance on robotic manipulation tasks with up to 92.8% success on RoboTwin and 78.3% on real-world tasks, while reaching 24.17 Hz closed-loop control.
WALL-WM advances video-action learning by using semantic events as learning units instead of fixed action chunks, enabling more flexible and scalable vision-language-action training and inference.
Introduces two projects related to robot world models: Awesome-WAM (OpenMOSS) includes papers such as World Action Models and DreamDojo; awesome-physical-ai curates a collection of papers on VLA models, world models, and embodied foundation models (including NVIDIA Cosmos Predict2.5).
NVIDIA's head of robotics, Jim Fan, gave a public talk, advocating that robots should directly replicate the successful path of large language models. He proposed directions such as World Action Model (WAM), a data revolution based on human first-person video, and neural simulation, and predicted a 95% probability of achieving the endgame of general-purpose physical robots by 2040.