DexHiL: A Human-in-the-Loop Framework for Vision-Language-Action Model Post-Training in Dexterous Manipulation

While Vision-Language-Action (VLA) models have demonstrated promising generalization capabilities in robotic manipulation, deploying them on specific and complex downstream tasks still demands effective post-training. In parallel, Human-in-the-Loop (HiL) learning has proven to be a powerful mechanism for refining robot policies. However, extending this paradigm to dexterous manipulation remains challenging: multi-finger control is high-dimensional, contact-intensive, and exhibits execution distributions that differ markedly from standard arm motions, leaving existing dexterous VLA systems limited in reliability and adaptability. We present DexHiL, the first integrated arm-hand human-in-the-loop framework for dexterous VLA models, enabling coordinated interventions over the arm and the dexterous hand within a single system. DexHiL introduces an intervention-aware data sampling strategy that prioritizes corrective segments for post-training, alongside a lightweight teleoperation interface that supports instantaneous human corrections during execution. Real-robot experiments demonstrate that DexHiL serves as an effective post-training framework, yielding a substantial performance leap, outperforming standard offline-only fine-tuning baselines by an average of 25% in success rates across distinct tasks. Project page: https://chenzhongxi-sjtu.github.io/dexhil/

Act, Sense, Act: Learning Non-Markovian Active Perception Strategies from Large-Scale Egocentric Human Data

Achieving generalizable manipulation in unconstrained environments requires the robot to proactively resolve information uncertainty, i.e., the capability of active perception. However, existing methods are often confined in limited types of sensing behaviors, restricting their applicability to complex environments. In this work, we formalize active perception as a non-Markovian process driven by information gain and decision branching, providing a structured categorization of visual active perception paradigms. Building on this perspective, we introduce CoMe-VLA, a cognitive and memory-aware vision-language-action (VLA) framework that leverages large-scale human egocentric data to learn versatile exploration and manipulation priors. Our framework integrates a cognitive auxiliary head for autonomous sub-task transitions and a dual-track memory system to maintain consistent self and environmental awareness by fusing proprioceptive and visual temporal contexts. By aligning human and robot hand-eye coordination behaviors in a unified egocentric action space, we train the model progressively in three stages. Extensive experiments on a wheel-based humanoid have demonstrated strong robustness and adaptability of our proposed method across diverse long-horizon tasks spanning multiple active perception scenarios.

Viewpoint Matters: Dynamically Optimizing Viewpoints with Masked Autoencoder for Visual Manipulation

Robotic manipulation continues to be a challenge, and imitation learning (IL) enables robots to learn tasks from expert demonstrations. Current IL methods typically rely on fixed camera setups, where cameras are manually positioned in static locations, imposing significant limitations on adaptability and coverage. Inspired by human active perception, where humans dynamically adjust their viewpoint to capture the most relevant and least noisy information, we propose MAE-Select, a novel framework for active viewpoint selection in single-camera robotic systems. MAE-Select fully leverages pre-trained multi-view masked autoencoder representations and dynamically selects the next most informative viewpoint at each time chunk without requiring labeled viewpoints. Extensive experiments demonstrate that MAE-Select improves the capabilities of single-camera systems and, in some cases, even surpasses multi-camera setups. The project will be available at https://mae-select.github.io.

EmboCoach-Bench: Benchmarking AI Agents on Developing Embodied Robots

The field of Embodied AI is witnessing a rapid evolution toward general-purpose robotic systems, fueled by high-fidelity simulation and large-scale data collection. However, this scaling capability remains severely bottlenecked by a reliance on labor-intensive manual oversight from intricate reward shaping to hyperparameter tuning across heterogeneous backends. Inspired by LLMs' success in software automation and science discovery, we introduce \textsc{EmboCoach-Bench}, a benchmark evaluating the capacity of LLM agents to autonomously engineer embodied policies. Spanning 32 expert-curated RL and IL tasks, our framework posits executable code as the universal interface. We move beyond static generation to assess a dynamic closed-loop workflow, where agents leverage environment feedback to iteratively draft, debug, and optimize solutions, spanning improvements from physics-informed reward design to policy architectures such as diffusion policies. Extensive evaluations yield three critical insights: (1) autonomous agents can qualitatively surpass human-engineered baselines by 26.5\% in average success rate; (2) agentic workflow with environment feedback effectively strengthens policy development and substantially narrows the performance gap between open-source and proprietary models; and (3) agents exhibit self-correction capabilities for pathological engineering cases, successfully resurrecting task performance from near-total failures through iterative simulation-in-the-loop debugging. Ultimately, this work establishes a foundation for self-evolving embodied intelligence, accelerating the paradigm shift from labor-intensive manual tuning to scalable, autonomous engineering in embodied AI field.

FSAG: Enhancing Human-to-Dexterous-Hand Finger-Specific Affordance Grounding via Diffusion Models

Dexterous grasp synthesis remains a central challenge: the high dimensionality and kinematic diversity of multi-fingered hands prevent direct transfer of algorithms developed for parallel-jaw grippers. Existing approaches typically depend on large, hardware-specific grasp datasets collected in simulation or through costly real-world trials, hindering scalability as new dexterous hand designs emerge. To this end, we propose a data-efficient framework, which is designed to bypass robot grasp data collection by exploiting the rich, object-centric semantic priors latent in pretrained generative diffusion models. Temporally aligned and fine-grained grasp affordances are extracted from raw human video demonstrations and fused with 3D scene geometry from depth images to infer semantically grounded contact targets. A kinematics-aware retargeting module then maps these affordance representations to diverse dexterous hands without per-hand retraining. The resulting system produces stable, functionally appropriate multi-contact grasps that remain reliably successful across common objects and tools, while exhibiting strong generalization across previously unseen object instances within a category, pose variations, and multiple hand embodiments. This work (i) introduces a semantic affordance extraction pipeline leveraging vision-language generative priors for dexterous grasping, (ii) demonstrates cross-hand generalization without constructing hardware-specific grasp datasets, and (iii) establishes that a single depth modality suffices for high-performance grasp synthesis when coupled with foundation-model semantics. Our results highlight a path toward scalable, hardware-agnostic dexterous manipulation driven by human demonstrations and pretrained generative models.

Look, Zoom, Understand: The Robotic Eyeball for Embodied Perception

In embodied AI perception systems, visual perception should be active: the goal is not to passively process static images, but to actively acquire more informative data within pixel and spatial budget constraints. Existing vision models and fixed RGB-D camera systems fundamentally fail to reconcile wide-area coverage with fine-grained detail acquisition, severely limiting their efficacy in open-world robotic applications. To address this issue, we propose EyeVLA, a robotic eyeball for active visual perception that can take proactive actions based on instructions, enabling clear observation of fine-grained target objects and detailed information across a wide spatial extent. EyeVLA discretizes action behaviors into action tokens and integrates them with vision-language models (VLMs) that possess strong open-world understanding capabilities, enabling joint modeling of vision, language, and actions within a single autoregressive sequence. By using the 2D bounding box coordinates to guide the reasoning chain and applying reinforcement learning to refine the viewpoint selection policy, we transfer the open-world scene understanding capability of the VLM to a vision language action (VLA) policy using only minimal real-world data. Experiments show that our system efficiently performs instructed scenes in real-world environments and actively acquires more accurate visual information through instruction-driven actions of rotation and zoom, thereby achieving strong environmental perception capabilities. EyeVLA introduces a novel robotic vision system that leverages detailed and spatially rich, large-scale embodied data, and actively acquires highly informative visual observations for downstream embodied tasks.

SAGE: Scene Graph-Aware Guidance and Execution for Long-Horizon Manipulation Tasks

Successfully solving long-horizon manipulation tasks remains a fundamental challenge. These tasks involve extended action sequences and complex object interactions, presenting a critical gap between high-level symbolic planning and low-level continuous control. To bridge this gap, two essential capabilities are required: robust long-horizon task planning and effective goal-conditioned manipulation. Existing task planning methods, including traditional and LLM-based approaches, often exhibit limited generalization or sparse semantic reasoning. Meanwhile, image-conditioned control methods struggle to adapt to unseen tasks. To tackle these problems, we propose SAGE, a novel framework for Scene Graph-Aware Guidance and Execution in Long-Horizon Manipulation Tasks. SAGE utilizes semantic scene graphs as a structural representation for scene states. A structural scene graph enables bridging task-level semantic reasoning and pixel-level visuo-motor control. This also facilitates the controllable synthesis of accurate, novel sub-goal images. SAGE consists of two key components: (1) a scene graph-based task planner that uses VLMs and LLMs to parse the environment and reason about physically-grounded scene state transition sequences, and (2) a decoupled structural image editing pipeline that controllably converts each target sub-goal graph into a corresponding image through image inpainting and composition. Extensive experiments have demonstrated that SAGE achieves state-of-the-art performance on distinct long-horizon tasks.

FoAM: Foresight-Augmented Multi-Task Imitation Policy for Robotic Manipulation

Multi-task imitation learning (MTIL) has shown significant potential in robotic manipulation by enabling agents to perform various tasks using a unified policy. This simplifies the policy deployment and enhances the agent's adaptability across different contexts. However, key challenges remain, such as maintaining action reliability (e.g., avoiding abnormal action sequences that deviate from nominal task trajectories), distinguishing between similar tasks, and generalizing to unseen scenarios. To address these challenges, we introduce the Foresight-Augmented Manipulation Policy (FoAM), an innovative MTIL framework. FoAM not only learns to mimic expert actions but also predicts the visual outcomes of those actions to enhance decision-making. Additionally, it integrates multi-modal goal inputs, such as visual and language prompts, overcoming the limitations of single-conditioned policies. We evaluated FoAM across over 100 tasks in both simulation and real-world settings, demonstrating that it significantly improves IL policy performance, outperforming current state-of-the-art IL baselines by up to 41% in success rate. Furthermore, we released a simulation benchmark for robotic manipulation, featuring 10 task suites and over 80 challenging tasks designed for multi-task policy training and evaluation. See project homepage https://projFoAM.github.io/ for project details.

Cable Routing and Assembly using Tactile-driven Motion Primitives

Manipulating cables is challenging for robots because of the infinite degrees of freedom of the cables and frequent occlusion by the gripper and the environment. These challenges are further complicated by the dexterous nature of the operations required for cable routing and assembly, such as weaving and inserting, hampering common solutions with vision-only sensing. In this paper, we propose to integrate tactile-guided low-level motion control with high-level vision-based task parsing for a challenging task: cable routing and assembly on a reconfigurable task board. Specifically, we build a library of tactile-guided motion primitives using a fingertip GelSight sensor, where each primitive reliably accomplishes an operation such as cable following and weaving. The overall task is inferred via visual perception given a goal configuration image, and then used to generate the primitive sequence. Experiments demonstrate the effectiveness of individual tactile-guided primitives and the integrated end-to-end solution, significantly outperforming the method without tactile sensing. Our reconfigurable task setup and proposed baselines provide a benchmark for future research in cable manipulation. More details and video are presented in \url{https://helennn.github.io/cable-manip/}

Prim-LAfD: A Framework to Learn and Adapt Primitive-Based Skills from Demonstrations for Insertion Tasks

Learning generalizable insertion skills in a data-efficient manner has long been a challenge in the robot learning community. While the current state-of-the-art methods with reinforcement learning (RL) show promising performance in acquiring manipulation skills, the algorithms are data-hungry and hard to generalize. To overcome the issues, in this paper we present Prim-LAfD, a simple yet effective framework to learn and adapt primitive-based insertion skills from demonstrations. Prim-LAfD utilizes black-box function optimization to learn and adapt the primitive parameters leveraging prior experiences. Human demonstrations are modeled as dense rewards guiding parameter learning. We validate the effectiveness of the proposed method on eight peg-hole and connector-socket insertion tasks. The experimental results show that our proposed framework takes less than one hour to acquire the insertion skills and as few as fifteen minutes to adapt to an unseen insertion task on a physical robot.