UniWAM: Unified World-Action Model

Hugging Face Daily Papers Papers
vision-language-action world-model robotics embodied-ai flow-matching scaling-law

Summary

UniWAM 提出了一种统一的架构，将物理推理器、世界生成器和动作预测器整合在一起，联合学习物理世界的语义理解、视觉生成与动作预测。该模型在多项评估中达到 SOTA，并揭示了统一人机协同训练的对数线性缩放规律。

Vision-language-action models benefit from the understanding and reasoning capabilities of pretrained vision-language models, but action-only supervision provides limited grounding in world dynamics. Conversely, world-action models inherit spatiotemporal priors from video generation models, yet remain limited in semantic understanding and reasoning under distribution shifts. We introduce UniWAM, a unified architecture that integrates a physical reasoner, a world generator, and an action predictor to jointly learn semantic understanding of the physical world, visual generation, and action prediction. To ensure the quality of the training data, we developed a rigorous data cleaning and annotation pipeline for both human egocentric data and robot data. To adapt the vision-language component to embodied tasks while preserving its inherited language capabilities, we represent low-level actions in natural language and introduce a pre-training recipe that assigns complementary supervision from visual question answering (VQA) data, human egocentric data, and robot demonstrations to the appropriate model components. During post-training, future visual noise augmentation reduces reliance on precise future predictions, while history-conditioned flow matching uses encoded action history to initialize action generation. Together, these designs significantly reduce denoising steps while maintaining performance. UniWAM achieves state-of-the-art (SOTA) performance across multiple evaluations, including in-distribution performance, robustness, generalization, instruction following, and long-horizon task execution. Furthermore, we uncover a log-linear scaling law of unified human-robot co-training, demonstrating the effectiveness of large-scale pre-training on a mixture of human and robot data.
Original Article
View Cached Full Text

Cached at: 10/08/26, 12:19 PM

Paper page - UniWAM: Unified World-Action Model

Source: https://huggingface.co/papers/2610.02054 Authors:

,

,

,

,

,

,

,

,

,

,

,

,

,

,

Abstract

Vision-language-actionmodelsbenefitfromtheunderstandingandreasoningcapabilitiesofpretrainedvision-languagemodels,butaction-onlysupervisionprovideslimitedgroundinginworlddynamics.Conversely,world-actionmodelsinheritspatiotemporalpriorsfromvideogenerationmodels,yetremainlimitedinsemanticunderstandingandreasoningunderdistributionshifts.WeintroduceUniWAM,aunifiedarchitecturethatintegratesaphysicalreasoner,aworldgenerator,andanactionpredictortojointlylearnsemanticunderstandingofthephysicalworld,visualgeneration,andactionprediction.Toensurethequalityofthetrainingdata,wedevelopedarigorousdatacleaningandannotationpipelineforbothhumanegocentricdataandrobotdata.Toadaptthevision-languagecomponenttoembodiedtaskswhilepreservingitsinheritedlanguagecapabilities,werepresentlow-levelactionsinnaturallanguageandintroduceapre-trainingrecipethatassignscomplementarysupervisionfromvisualquestionanswering(VQA)data,humanegocentricdata,androbotdemonstrationstotheappropriatemodelcomponents.Duringpost-training,futurevisualnoiseaugmentationreducesrelianceonprecisefuturepredictions,whilehistory-conditionedflowmatchingusesencodedactionhistorytoinitializeactiongeneration.Together,thesedesignssignificantlyreducedenoisingstepswhilemaintainingperformance.UniWAMachievesstate-of-the-art(SOTA)performanceacrossmultipleevaluations,includingin-distributionperformance,robustness,generalization,instructionfollowing,andlong-horizontaskexecution.Furthermore,weuncoveralog-linearscalinglawofunifiedhuman-robotco-training,demonstratingtheeffectivenessoflarge-scalepre-trainingonamixtureofhumanandrobotdata.

View arXiv pageView PDFProject pageGitHub40Add to collection

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2610.02054 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2610.02054 in a dataset README.md to link it from this page.

Spaces citing this paper1

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

UNITAS: A 3D-Native World Action Model for Embodied Manipulation

Hugging Face Daily Papers

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。

GreenWAM: A World Foundation Model for Robotics

Hugging Face Daily Papers

GreenWAM 提出一个机器人世界基础模型，通过共享的解剖学动作空间联合训练 20 个机器人家族，使其既能作为世界模型预测动作后果，又能作为策略控制不同形态的机器人，并在多个基准上取得领先表现。

AffordanceWAM: Affordance-Aware Joint World-Action Modeling for Robot Manipulation

Hugging Face Daily Papers

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。

Long-WAM: Scaling the Context of World-Action Models

Hugging Face Daily Papers

Long-WAM 提出了一种在实时控制约束下扩展因果世界-动作模型上下文的模型-系统框架，发现更长的视觉历史仅在视频基础模型以自回归方式预训练时才显著有效，在 RoboCasa GR-1 上将成功率从 63.3% 提升至 78.7%，并在 Unitree G1 实机上实现 95% 的动态杯子堆叠成功率。