arXiv subjects

Zongqing Lu

Publications and source records attributed to Zongqing Lu.

At least 37 records · Page 2Linked to original sources

arXiv2026

Guided Policy Optimization under Partial Observability

Reinforcement Learning (RL) in partially observable environments poses significant challenges due to the complexity of learning under uncertainty. While additional information, such as that available in simulations, can enhance training, effectively leveraging it remains an open problem. To address this, we introduce Guided Policy Optimization (GPO), a framework that co-trains a guider and a learner. The guider takes advantage of privileged information while ensuring alignment with the learner's policy that is primarily trained via imitation learning. We theoretically demonstrate that this learning scheme achieves optimality comparable to direct RL, thereby overcoming key limitations inherent in existing approaches. Empirical evaluations show strong performance of GPO across various tasks, including continuous control with partial observability and noise, and memory-based challenges, significantly outperforming existing methods.

cs.LG↗

arXiv2026

Multi-Agent Guided Policy Optimization

Due to practical constraints such as partial observability and limited communication, Centralized Training with Decentralized Execution (CTDE) has become the dominant paradigm in cooperative Multi-Agent Reinforcement Learning (MARL). However, existing CTDE methods often underutilize centralized training or lack theoretical guarantees. We propose Multi-Agent Guided Policy Optimization (MAGPO), a novel framework that better leverages centralized training by integrating centralized guidance with decentralized execution. MAGPO uses an autoregressive joint policy for scalable, coordinated exploration and explicitly aligns it with decentralized policies to ensure deployability under partial observability. We provide theoretical guarantees of monotonic policy improvement and empirically evaluate MAGPO on 43 tasks across 6 diverse environments. Results show that MAGPO consistently outperforms strong CTDE baselines and matches or surpasses fully centralized approaches, offering a principled and practical solution for decentralized multi-agent learning. Our code and experimental data can be found in https://github.com/liyheng/MAGPO.

cs.AI↗

arXiv2026

GTR-Turbo: Merged Checkpoint is Secretly a Free Teacher for Agentic VLM Training

Multi-turn reinforcement learning (RL) for multi-modal agents built upon vision-language models (VLMs) is hampered by sparse rewards and long-horizon credit assignment. Recent methods densify the reward by querying a teacher that provides step-level feedback, e.g., Guided Thought Reinforcement (GTR) and On-Policy Distillation, but rely on costly, often privileged models as the teacher, limiting practicality and reproducibility. We introduce GTR-Turbo, a highly efficient upgrade to GTR that matches its performance without training on or querying an expensive teacher model. Specifically, GTR-Turbo merges the weights of checkpoints produced during ongoing RL training and then uses the resulting merged model as a "free" teacher to guide subsequent RL via supervised fine-tuning or soft logit distillation. This design removes dependence on privileged VLMs (e.g., GPT or Gemini), mitigates the "entropy collapse" observed in prior work, and maintains stable training. Across diverse visual agentic tasks, GTR-Turbo improves the accuracy of the baseline model by 10-30% while reducing wall-clock training time by 50% and compute cost by 60% relative to GTR.

cs.CV↗

arXiv2026

Joint-Aligned Latent Action: Towards Scalable VLA Pretraining in the Wild

Despite progress, Vision-Language-Action models (VLAs) are limited by a scarcity of large-scale, diverse robot data. While human manipulation videos offer a rich alternative, existing methods are forced to choose between small, precisely-labeled datasets and vast in-the-wild footage with unreliable hand tracking labels. We present JALA, a pretraining framework that learns Jointly-Aligned Latent Actions. JALA bypasses full visual dynamic reconstruction, instead learns a predictive action embedding aligned with both inverse dynamics and real actions. This yields a transition-aware, behavior-centric latent space for learning from heterogeneous human data. We scale this approach with UniHand-Mix, a 7.5M video corpus (>2,000 hours) blending laboratory and in-the-wild footage. Experiments demonstrate that JALA generates more realistic hand motions in both controlled and unconstrained scenarios, significantly improving downstream robot manipulation performance in both simulation and real-world tasks. These results indicate that jointly-aligned latent actions offer a scalable pathway for VLA pretraining from human data.

cs.RO↗

arXiv2026

Temporal Difference Learning with Constrained Initial Representations

Recently, there have been numerous attempts to enhance the sample efficiency of off-policy reinforcement learning (RL) agents when interacting with the environment, including architecture improvements and new algorithms. Despite these advances, they overlook the potential of directly constraining the initial representations of the input data, which can intuitively alleviate the distribution shift issue and stabilize training. In this paper, we introduce the Tanh function into the initial layer to fulfill such a constraint. We theoretically unpack the convergence property of the temporal difference learning with the Tanh function under linear function approximation. Motivated by theoretical insights, we present our Constrained Initial Representations framework, tagged CIR, which is made up of three components: (i) the Tanh activation along with normalization methods to stabilize representations; (ii) the skip connection module to provide a linear pathway from the shallow layer to the deep layer; (iii) the convex Q-learning that allows a more flexible value estimate and mitigates potential conservatism. Empirical results show that CIR exhibits strong performance on numerous continuous control tasks, even being competitive or surpassing existing strong baseline methods.

cs.LG↗

arXiv2026

General Humanoid Whole-Body Control via Pretraining and Fast Adaptation

Learning a general whole-body controller for humanoid robots remains challenging due to the diversity of motion distributions, the difficulty of fast adaptation, and the need for robust balance in high-dynamic scenarios. Existing approaches often require task-specific training or suffer from performance degradation when adapting to new motions. In this paper, we present FAST, a general humanoid whole-body control framework that enables Fast Adaptation and Stable Motion Tracking. FAST introduces Parseval-Guided Residual Policy Adaptation, which learns a lightweight delta action policy under orthogonality and KL constraints, enabling efficient adaptation to out-of-distribution motions while mitigating catastrophic forgetting. To further improve physical robustness, we propose Center-of-Mass-Aware Control, which incorporates CoM-related observations and objectives to enhance balance when tracking challenging reference motions. Extensive experiments in simulation and real-world deployment demonstrate that FAST consistently outperforms state-of-the-art baselines in robustness, adaptation efficiency, and generalization.

cs.RO↗

arXiv2026

Rethinking Visual-Language-Action Model Scaling: Alignment, Mixture, and Regularization

While Vision-Language-Action (VLA) models show strong promise for generalist robot control, it remains unclear whether -- and under what conditions -- the standard "scale data" recipe translates to robotics, where training data is inherently heterogeneous across embodiments, sensors, and action spaces. We present a systematic, controlled study of VLA scaling that revisits core training choices for pretraining across diverse robots. Using a representative VLA framework that combines a vision-language backbone with flow-matching, we ablate key design decisions under matched conditions and evaluate in extensive simulation and real-robot experiments. To improve the reliability of real-world results, we introduce a Grouped Blind Ensemble protocol that blinds operators to model identity and separates policy execution from outcome judgment, reducing experimenter bias. Our analysis targets three dimensions of VLA scaling. (1) Physical alignment: we show that a unified end-effector (EEF)-relative action representation is critical for robust cross-embodiment transfer. (2) Embodiment mixture: we find that naively pooling heterogeneous robot datasets often induces negative transfer rather than gains, underscoring the fragility of indiscriminate data scaling. (3) Training regularization: we observe that intuitive strategies, such as sensory dropout and multi-stage fine-tuning, do not consistently improve performance at scale. Together, this study challenge some common assumptions about embodied scaling and provide practical guidance for training large-scale VLA policies from diverse robotic data. Project website: https://research.beingbeyond.com/rethink_vla

cs.RO↗

arXiv2026

Being-H0.5: Scaling Human-Centric Robot Learning for Cross-Embodiment Generalization

We introduce Being-H0.5, a foundational Vision-Language-Action (VLA) model designed for robust cross-embodiment generalization across diverse robotic platforms. While existing VLAs often struggle with morphological heterogeneity and data scarcity, we propose a human-centric learning paradigm that treats human interaction traces as a universal "mother tongue" for physical interaction. To support this, we present UniHand-2.0, the largest embodied pre-training recipe to date, comprising over 35,000 hours of multimodal data across 30 distinct robotic embodiments. Our approach introduces a Unified Action Space that maps heterogeneous robot controls into semantically aligned slots, enabling low-resource robots to bootstrap skills from human data and high-resource platforms. Built upon this human-centric foundation, we design a unified sequential modeling and multi-task pre-training paradigm to bridge human demonstrations and robotic execution. Architecturally, Being-H0.5 utilizes a Mixture-of-Transformers design featuring a novel Mixture-of-Flow (MoF) framework to decouple shared motor primitives from specialized embodiment-specific experts. Finally, to make cross-embodiment policies stable in the real world, we introduce Manifold-Preserving Gating for robustness under sensory shift and Universal Async Chunking to universalize chunked control across embodiments with different latency and control profiles. We empirically demonstrate that Being-H0.5 achieves state-of-the-art results on simulated benchmarks, such as LIBERO (98.9%) and RoboCasa (53.9%), while also exhibiting strong cross-embodiment capabilities on five robotic platforms.

cs.RO↗

arXiv2026

SmoothSync: Dual-Stream Diffusion Transformers for Jitter-Robust Beat-Synchronized Gesture Generation from Quantized Audio

Co-speech gesture generation is a critical area of research aimed at synthesizing speech-synchronized human-like gestures. Existing methods often suffer from issues such as rhythmic inconsistency, motion jitter, foot sliding and limited multi-sampling diversity. In this paper, we present SmoothSync, a novel framework that leverages quantized audio tokens in a novel dual-stream Diffusion Transformer (DiT) architecture to synthesis holistic gestures and enhance sampling variation. Specifically, we (1) fuse audio-motion features via complementary transformer streams to achieve superior synchronization, (2) introduce a jitter-suppression loss to improve temporal smoothness, (3) implement probabilistic audio quantization to generate distinct gesture sequences from identical inputs. To reliably evaluate beat synchronization under jitter, we introduce Smooth-BC, a robust variant of the beat consistency metric less sensitive to motion noise. Comprehensive experiments on the BEAT2 and SHOW datasets demonstrate SmoothSync's superiority, outperforming state-of-the-art methods by -30.6% FGD, 10.3% Smooth-BC, and 8.4% Diversity on BEAT2, while reducing jitter and foot sliding by -62.9% and -17.1% respectively. The code will be released to facilitate future research.

cs.SD↗

arXiv2025

UniTacHand: Unified Spatio-Tactile Representation for Human to Robotic Hand Skill Transfer

Tactile sensing is crucial for robotic hands to achieve human-level dexterous manipulation, especially in scenarios with visual occlusion. However, its application is often hindered by the difficulty of collecting large-scale real-world robotic tactile data. In this study, we propose to collect low-cost human manipulation data using haptic gloves for tactile-based robotic policy learning. The misalignment between human and robotic tactile data makes it challenging to transfer policies learned from human data to robots. To bridge this gap, we propose UniTacHand, a unified representation to align robotic tactile information captured by dexterous hands with human hand touch obtained from gloves. First, we project tactile signals from both human hands and robotic hands onto a morphologically consistent 2D surface space of the MANO hand model. This unification standardizes the heterogeneous data structures and inherently embeds the tactile signals with spatial context. Then, we introduce a contrastive learning method to align them into a unified latent space, trained on only 10 minutes of paired data from our data collection system. Our approach enables zero-shot tactile-based policy transfer from humans to a real robot, generalizing to objects unseen in the pre-training data. We also demonstrate that co-training on mixed data, including both human and robotic demonstrations via UniTacHand, yields better performance and data efficiency compared with using only robotic data. UniTacHand paves a path toward general, scalable, and data-efficient learning for tactile-based dexterous hands.

cs.RO↗

arXiv2025

Creative Agents: Empowering Agents with Imagination for Creative Tasks

We study building embodied agents for open-ended creative tasks. While existing methods build instruction-following agents that can perform diverse open-ended tasks, none of them demonstrates creativity -- the ability to give novel and diverse solutions implicit in the language instructions. This limitation comes from their inability to convert abstract language instructions into concrete goals and perform long-horizon planning for such complicated goals. Given the observation that humans perform creative tasks with imagination, we propose a class of solutions, where the controller is enhanced with an imaginator generating detailed imaginations of task outcomes conditioned on language instructions. We introduce several approaches to implementing the components of creative agents. We implement the imaginator with either a large language model for textual imagination or a diffusion model for visual imagination. The controller can either be a behavior-cloning policy or a pre-trained foundation model generating executable codes in the environment. We benchmark creative tasks with the challenging open-world game Minecraft, where the agents create diverse buildings given free-form language instructions. We propose novel evaluation metrics for open-ended creative tasks utilizing GPT-4V, which holds many advantages over existing metrics. We perform a detailed experimental analysis of creative agents, showing that creative agents are the first AI agents accomplishing diverse building creation in the survival mode of Minecraft. Our benchmark and models are open-source for future research on creative agents (https://github.com/PKU-RL/Creative-Agents).

cs.AI↗

arXiv2025

Spatial-Aware VLA Pretraining through Visual-Physical Alignment from Human Videos

Vision-Language-Action (VLA) models provide a promising paradigm for robot learning by integrating visual perception with language-guided policy learning. However, most existing approaches rely on 2D visual inputs to perform actions in 3D physical environments, creating a significant gap between perception and action grounding. To bridge this gap, we propose a Spatial-Aware VLA Pretraining paradigm that performs explicit alignment between visual space and physical space during pretraining, enabling models to acquire 3D spatial understanding before robot policy learning. Starting from pretrained vision-language models, we leverage large-scale human demonstration videos to extract 3D visual and 3D action annotations, forming a new source of supervision that aligns 2D visual observations with 3D spatial reasoning. We instantiate this paradigm with VIPA-VLA, a dual-encoder architecture that incorporates a 3D visual encoder to augment semantic visual representations with 3D-aware features. When adapted to downstream robot tasks, VIPA-VLA achieves significantly improved grounding between 2D vision and 3D action, resulting in more robust and generalizable robotic policies.

cs.RO↗

arXiv2025

Universal Dexterous Functional Grasping via Demonstration-Editing Reinforcement Learning

Reinforcement learning (RL) has achieved great success in dexterous grasping, significantly improving grasp performance and generalization from simulation to the real world. However, fine-grained functional grasping, which is essential for downstream manipulation tasks, remains underexplored and faces several challenges: the complexity of specifying goals and reward functions for functional grasps across diverse objects, the difficulty of multi-task RL exploration, and the challenge of sim-to-real transfer. In this work, we propose DemoFunGrasp for universal dexterous functional grasping. We factorize functional grasping conditions into two complementary components - grasping style and affordance - and integrate them into an RL framework that can learn to grasp any object with any functional grasping condition. To address the multi-task optimization challenge, we leverage a single grasping demonstration and reformulate the RL problem as one-step demonstration editing, substantially enhancing sample efficiency and performance. Experimental results in both simulation and the real world show that DemoFunGrasp generalizes to unseen combinations of objects, affordances, and grasping styles, outperforming baselines in both success rate and functional grasping accuracy. In addition to strong sim-to-real capability, by incorporating a vision-language model (VLM) for planning, our system achieves autonomous instruction-following grasp execution.

cs.RO↗

arXiv2025

SENTINEL: A Fully End-to-End Language-Action Model for Humanoid Whole Body Control

Existing humanoid control systems often rely on teleoperation or modular generation pipelines that separate language understanding from physical execution. However, the former is entirely human-driven, and the latter lacks tight alignment between language commands and physical behaviors. In this paper, we present SENTINEL, a fully end-to-end language-action model for humanoid whole-body control. We construct a large-scale dataset by tracking human motions in simulation using a pretrained whole body controller, combined with their text annotations. The model directly maps language commands and proprioceptive inputs to low-level actions without any intermediate representation. The model generates action chunks using flow matching, which can be subsequently refined by a residual action head for real-world deployment. Our method exhibits strong semantic understanding and stable execution on humanoid robots in both simulation and real-world deployment, and also supports multi-modal extensions by converting inputs into texts.

cs.RO↗

arXiv2025

Towards Proprioception-Aware Embodied Planning for Dual-Arm Humanoid Robots

In recent years, Multimodal Large Language Models (MLLMs) have demonstrated the ability to serve as high-level planners, enabling robots to follow complex human instructions. However, their effectiveness, especially in long-horizon tasks involving dual-arm humanoid robots, remains limited. This limitation arises from two main challenges: (i) the absence of simulation platforms that systematically support task evaluation and data collection for humanoid robots, and (ii) the insufficient embodiment awareness of current MLLMs, which hinders reasoning about dual-arm selection logic and body positions during planning. To address these issues, we present DualTHOR, a new dual-arm humanoid simulator, with continuous transition and a contingency mechanism. Building on this platform, we propose Proprio-MLLM, a model that enhances embodiment awareness by incorporating proprioceptive information with motion-based position embedding and a cross-spatial encoder. Experiments show that, while existing MLLMs struggle in this environment, Proprio-MLLM achieves an average improvement of 19.75% in planning performance. Our work provides both an essential simulation platform and an effective model to advance embodied intelligence in humanoid robotics. The code is available at https://anonymous.4open.science/r/DualTHOR-5F3B.

cs.RO↗

arXiv2025

DualTHOR: A Dual-Arm Humanoid Simulation Platform for Contingency-Aware Planning

Developing embodied agents capable of performing complex interactive tasks in real-world scenarios remains a fundamental challenge in embodied AI. Although recent advances in simulation platforms have greatly enhanced task diversity to train embodied Vision Language Models (VLMs), most platforms rely on simplified robot morphologies and bypass the stochastic nature of low-level execution, which limits their transferability to real-world robots. To address these issues, we present a physics-based simulation platform DualTHOR for complex dual-arm humanoid robots, built upon an extended version of AI2-THOR. Our simulator includes real-world robot assets, a task suite for dual-arm collaboration, and inverse kinematics solvers for humanoid robots. We also introduce a contingency mechanism that incorporates potential failures through physics-based low-level execution, bridging the gap to real-world scenarios. Our simulator enables a more comprehensive evaluation of the robustness and generalization of VLMs in household environments. Extensive evaluations reveal that current VLMs struggle with dual-arm coordination and exhibit limited robustness in realistic environments with contingencies, highlighting the importance of using our simulator to develop more capable VLMs for embodied tasks. The code is available at https://github.com/ds199895/DualTHOR.git.

cs.RO↗

arXiv2025

DemoHLM: From One Demonstration to Generalizable Humanoid Loco-Manipulation

Loco-manipulation is a fundamental challenge for humanoid robots to achieve versatile interactions in human environments. Although recent studies have made significant progress in humanoid whole-body control, loco-manipulation remains underexplored and often relies on hard-coded task definitions or costly real-world data collection, which limits autonomy and generalization. We present DemoHLM, a framework for humanoid loco-manipulation that enables generalizable loco-manipulation on a real humanoid robot from a single demonstration in simulation. DemoHLM adopts a hierarchy that integrates a low-level universal whole-body controller with high-level manipulation policies for multiple tasks. The whole-body controller maps whole-body motion commands to joint torques and provides omnidirectional mobility for the humanoid robot. The manipulation policies, learned in simulation via our data generation and imitation learning pipeline, command the whole-body controller with closed-loop visual feedback to execute challenging loco-manipulation tasks. Experiments show a positive correlation between the amount of synthetic data and policy performance, underscoring the effectiveness of our data generation pipeline and the data efficiency of our approach. Real-world experiments on a Unitree G1 robot equipped with an RGB-D camera validate the sim-to-real transferability of DemoHLM, demonstrating robust performance under spatial variations across ten loco-manipulation tasks.

cs.RO↗

arXiv2025

DemoGrasp: Universal Dexterous Grasping from a Single Demonstration

Universal grasping with multi-fingered dexterous hands is a fundamental challenge in robotic manipulation. While recent approaches successfully learn closed-loop grasping policies using reinforcement learning (RL), the inherent difficulty of high-dimensional, long-horizon exploration necessitates complex reward and curriculum design, often resulting in suboptimal solutions across diverse objects. We propose DemoGrasp, a simple yet effective method for learning universal dexterous grasping. We start from a single successful demonstration trajectory of grasping a specific object and adapt to novel objects and poses by editing the robot actions in this trajectory: changing the wrist pose determines where to grasp, and changing the hand joint angles determines how to grasp. We formulate this trajectory editing as a single-step Markov Decision Process (MDP) and use RL to optimize a universal policy across hundreds of objects in parallel in simulation, with a simple reward consisting of a binary success term and a robot-table collision penalty. In simulation, DemoGrasp achieves a 95% success rate on DexGraspNet objects using the Shadow Hand, outperforming previous state-of-the-art methods. It also shows strong transferability, achieving an average success rate of 84.6% across diverse dexterous hand embodiments on six unseen object datasets, while being trained on only 175 objects. Through vision-based imitation learning, our policy successfully grasps 110 unseen real-world objects, including small, thin items. It generalizes to spatial, background, and lighting changes, supports both RGB and depth inputs, and extends to language-guided grasping in cluttered scenes.

cs.RO↗