Xinyi Mao

Yao Feng, Chendong Xiang, Xinyi Mao et al.

Robotic arm manipulation in data-scarce settings is a highly challenging task due to the complex embodiment dynamics and diverse contexts. Recent video-based approaches have shown great promise in capturing and transferring the temporal and physical interactions by pre-training on Internet-scale video data. However, such methods are often not optimized for the embodiment-specific closed-loop control, typically suffering from high latency and insufficient grounding. In this paper, we present Vidarc (Video Diffusion for Action Reasoning and Closed-loop Control), a novel autoregressive embodied video diffusion approach augmented by a masked inverse dynamics model. By grounding video predictions with action-relevant masks and incorporating real-time feedback through cached autoregressive generation, Vidarc achieves fast, accurate closed-loop control. Pre-trained on one million cross-embodiment episodes, Vidarc surpasses state-of-the-art baselines, achieving at least a 15% higher success rate in real-world deployment and a 91% reduction in latency. We also highlight its robust generalization and error correction capabilities across previously unseen robotic platforms.

Guo Chen, Lidong Lu, Yicheng Liu et al.

While datasets for video understanding have scaled to hour-long durations, they typically consist of densely concatenated clips that differ from natural, unscripted daily life. To bridge this gap, we introduce MM-Lifelong, a dataset designed for Multimodal Lifelong Understanding. Comprising 181.1 hours of footage, it is structured across Day, Week, and Month scales to capture varying temporal densities. Extensive evaluations reveal two critical failure modes in current paradigms: end-to-end MLLMs suffer from a Working Memory Bottleneck due to context saturation, while representative agentic baselines experience Global Localization Collapse when navigating sparse, month-long timelines. To address this, we propose the Recursive Multimodal Agent (ReMA), which employs dynamic memory management to iteratively update a recursive belief state, significantly outperforming existing methods. Finally, we establish dataset splits designed to isolate temporal and domain biases, providing a rigorous foundation for future research in supervised learning and out-of-distribution generalization.

Hengkai Tan, Xuezhou Xu, Chengyang Ying et al. · tsinghua

Learning a precise robotic grasping policy is crucial for embodied agents operating in complex real-world manipulation tasks. Despite significant advancements, most models still struggle with accurate spatial positioning of objects to be grasped. We first show that this spatial generalization challenge stems primarily from the extensive data requirements for adequate spatial understanding. However, collecting such data with real robots is prohibitively expensive, and relying on simulation data often leads to visual generalization gaps upon deployment. To overcome these challenges, we then focus on state-based policy generalization and present \textbf{ManiBox}, a novel bounding-box-guided manipulation method built on a simulation-based teacher-student framework. The teacher policy efficiently generates scalable simulation data using bounding boxes, which are proven to uniquely determine the objects' spatial positions. The student policy then utilizes these low-dimensional spatial states to enable zero-shot transfer to real robots. Through comprehensive evaluations in simulated and real-world environments, ManiBox demonstrates a marked improvement in spatial grasping generalization and adaptability to diverse objects and backgrounds. Further, our empirical study into scaling laws for policy performance indicates that spatial volume generalization scales with data volume in a power law. For a certain level of spatial volume, the success rate of grasping empirically follows Michaelis-Menten kinetics relative to data volume, showing a saturation effect as data increases. Our videos and code are available in https://thkkk.github.io/manibox.

Yao Feng, Hengkai Tan, Xinyi Mao et al.

Scaling general-purpose manipulation to new robot embodiments remains challenging: each platform typically needs large, homogeneous demonstrations, and pixel-to-action VLA pipelines typically degenerate under background and viewpoint shifts. In this paper, we present Vidar, a prior-driven, low-shot adaptation paradigm that replaces most embodiment-specific data with transferable video priors. Vidar consists of an embodied video diffusion model as the generalizable prior and a masked inverse dynamics model (MIDM) adapter based on a key decoupling of the policy. The embodied diffusion model is pre-trained on Internet-scale videos and then domain-adapted to 750K multi-view trajectories from three real-world robot platforms using a unified observation space encoding robot, camera, task, and scene contexts. The MIDM module learns action-relevant pixel masks without dense labels, grounding the prior into the target embodiment's action space while suppressing distractors. Crucially, the generative video prior models the distribution of plausible, temporally coherent interactions, implicitly capturing affordances, contact dynamics, and physical consistency from massive unlabeled video. This shifts the challenge from collecting large amounts of new robot data to efficiently aligning a rich prior with a new embodiment. With only 20 minutes of human demonstrations on an unseen robot (1% of typical data), Vidar outperforms state-of-the-art VLA baselines and generalizes to unseen tasks, backgrounds, and camera layouts. Our results suggest a scalable recipe for "one prior, many embodiments": strong, inexpensive video priors + minimal on-robot alignment.

Hengkai Tan, Yao Feng, Xinyi Mao et al.

Vision-language-action (VLA) models have shown promise on task-conditioned control in complex settings such as bimanual manipulation. However, the heavy reliance on task-specific human demonstrations limits their generalization and incurs high data acquisition costs. In this work, we present a new notion of task-agnostic action paradigm that decouples action execution from task-specific conditioning, enhancing scalability, efficiency, and cost-effectiveness. To address the data collection challenges posed by this paradigm -- such as low coverage density, behavioral redundancy, and safety risks -- we introduce ATARA (Automated Task-Agnostic Random Actions), a scalable self-supervised framework that accelerates collection by over $ 30\times $ compared to human teleoperation. To further enable effective learning from task-agnostic data, which often suffers from distribution mismatch and irrelevant trajectories, we propose AnyPos, an inverse dynamics model equipped with Arm-Decoupled Estimation and a Direction-Aware Decoder (DAD). We additionally integrate a video-conditioned action validation module to verify the feasibility of learned policies across diverse manipulation tasks. Extensive experiments show that the AnyPos-ATARA pipeline yields a 51% improvement in test accuracy and achieves 30-40% higher success rates in downstream tasks such as lifting, pick-and-place, and clicking, using replay-based video validation. Project Page: https://embodiedfoundation.github.io/vidar_anypos