arXiv subjects

Zhigang Wang

Publications and source records attributed to Zhigang Wang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Hydrogen-atom roaming reactions in water clusters: Unveiling an unusual dimension of water reactivity through first-principles calculations and machine learning

Water mediates a broad range of chemical reactions, including proton transfer, bond rearrangement, and conventional radical processes, defining a continuously expanding repertoire of intrinsic reactivity. However, roaming, a fundamental reaction mechanism that a departing fragment bypasses the minimum energy path to recombine, has not been identified in water itself. Here, we report the discovery of hydrogen-atom roaming reactions in water clusters through high-precision ab initio calculations of first-principles. A neutral hydrogen atom departs as a radical, roams across the flat potential energy surface, and recombines along pathways that connect the same reactants and products as known hydrogen-bond network rearrangements. Interpretable machine learning analysis identifies the reactant dipole moment as the decisive switch governing whether roaming occurs, underpinned by exchange-repulsion and electrostatic interactions. Once roaming is initiated, polarizability and spin population determine barrier heights, while the charge distribution of the roaming hydrogen atom governs barrier widths, collectively shaped by electrostatic, orbital, and dispersion contributions. These findings establish hydrogen-atom roaming as a previously unrecognized intrinsic reaction class in water, complementing a fundamental dimension to the mechanistic picture of water reactivity.

physics.chem-ph↗

arXiv2026

Openfly: A comprehensive platform for aerial vision-language navigation

Vision-Language Navigation (VLN) aims to guide agents by leveraging language instructions and visual cues, playing a pivotal role in embodied AI. Indoor VLN has been extensively studied, whereas outdoor aerial VLN remains underexplored. The potential reason is that outdoor aerial view encompasses vast areas, making data collection more challenging, which results in a lack of benchmarks. To address this problem, we propose OpenFly, a platform comprising various rendering engines, a versatile toolchain, and a large-scale benchmark for aerial VLN. Firstly, we integrate diverse rendering engines and advanced techniques for environment simulation, including Unreal Engine, GTA V, Google Earth, and 3D Gaussian Splatting (3D GS). Particularly, 3D GS supports real-to-sim rendering, further enhancing the realism of our environments. Secondly, we develop a highly automated toolchain for aerial VLN data collection, streamlining point cloud acquisition, scene semantic segmentation, flight trajectory creation, and instruction generation. Thirdly, based on the toolchain, we construct a large-scale aerial VLN dataset with 100k trajectories, covering diverse heights and lengths across 18 scenes. Moreover, we propose OpenFly-Agent, a keyframe-aware VLN model emphasizing key observations during flight. For benchmarking, extensive experiments and analyses are conducted, evaluating several recent VLN methods and showcasing the superiority of our OpenFly platform and agent. The toolchain, dataset, and codes will be open-sourced.

cs.CV↗

arXiv2026

Understanding Degradation with Vision Language Model

Understanding visual degradations is a critical yet challenging problem in computer vision. While recent Vision-Language Models (VLMs) excel at qualitative description, they often fall short in understanding the parametric physics underlying image degradations. In this work, we redefine degradation understanding as a hierarchical structured prediction task, necessitating the concurrent estimation of degradation types, parameter keys, and their continuous physical values. Although these sub-tasks operate in disparate spaces, we prove that they can be unified under one autoregressive next-token prediction paradigm, whose error is bounded by the value-space quantization grid. Building on this insight, we introduce DU-VLM, a multimodal chain-of-thought model trained with supervised fine-tuning and reinforcement learning using structured rewards. Furthermore, we show that DU-VLM can serve as a zero-shot controller for pre-trained diffusion models, enabling high-fidelity image restoration without fine-tuning the generative backbone. We also introduce \textbf{DU-110k}, a large-scale dataset comprising 110,000 clean-degraded pairs with grounded physical annotations. Extensive experiments demonstrate that our approach significantly outperforms generalist baselines in both accuracy and robustness, exhibiting generalization to unseen distributions.

cs.CV↗

arXiv2025

Metallic solid-state hydrogen storage crystals achieved through chemical precompression under ambient conditions

Improving hydrogen storage density is essential for reducing the extreme conditions required in applications such as nuclear fusion. However, the recognition of metallic hydrogen as the "Holy Grail" of high-pressure science highlights the difficulty of high-density hydrogen aggregation. Here, we report a solid-state crystal H9@C20 formed by embedding hydrogen atoms into C20 fullerene cages and utilizing chemical precompression, which remains stable under ambient pressure and temperature conditions and exhibits metallic properties. This precompression effect is reflected in the formation of C-H bonds within the cage and C-C bonds between cages, resulting in the transformation of all C atoms from sp2 to sp3 hybridization with inward and outward distortions, while promoting delocalized multicenter bonding within the H9 aggregate. In particular, the hydrogen density inside the C20 cage exceeds that of solid hydrogen, achieving a uniform discrete distribution with H9 as monomers. Further study reveals that filling hydrogen molecules into voids between H9@C20 primitive cells can increase hydrogen content while maintaining structural stability, forming a solid-gas mixed hydrogen storage crystal. Our findings provide a basis for developing high-density hydrogen storage materials under ambient conditions.

cond-mat.mtrl-sci↗

arXiv2025

Exploring the Potential of Encoder-free Architectures in 3D LMMs

Encoder-free architectures have been preliminarily explored in the 2D Large Multimodal Models (LMMs), yet it remains an open question whether they can be effectively applied to 3D understanding scenarios. In this paper, we present the first comprehensive investigation into the potential of encoder-free architectures to alleviate the challenges of encoder-based 3D LMMs. These long-standing challenges include the failure to adapt to varying point cloud resolutions during inference and the point features from the encoder not meeting the semantic needs of Large Language Models (LLMs). We identify key aspects for 3D LMMs to remove the pre-trained encoder and enable the LLM to assume the role of the 3D encoder: 1) We propose the LLM-embedded Semantic Encoding strategy in the pre-training stage, exploring the effects of various point cloud self-supervised losses. And we present the Hybrid Semantic Loss to extract high-level semantics. 2) We introduce the Hierarchical Geometry Aggregation strategy in the instruction tuning stage. This incorporates inductive bias into the LLM layers to focus on the local details of the point clouds. To the end, we present the first Encoder-free 3D LMM, ENEL. Our 7B model rivals the state-of-the-art model, PointLLM-PiSA-13B, achieving 57.91%, 61.0%, and 55.20% on the classification, captioning, and VQA tasks, respectively. Our results show that the encoder-free architecture is highly promising for replacing encoder-based architectures in the field of 3D understanding. The code is released at https://github.com/Ivan-Tang-3D/ENEL

cs.CV↗

arXiv2025

A Model and Data Dual-driven Approach for Multitargets Detection under Mainlobe Jamming

In modern radar systems, target detection and parameter estimation face significant challenges when confronted with mainlobe jamming. This paper presents a Diffusion-based Model and Data Dual-driven (DMDD) approach to estimate and detect multitargets and suppress structured jamming. In DMDD, the jamming prior is modeled through a score-based diffusion process with its score learned from the pure jamming data, enabling posterior sampling without requiring detailed knowledge of jamming. Meanwhile, the target signal is usually sparse in the range space, which can be modeled via a sparse Bayesian learning (SBL) framework, and hyperparameter is updated through the expectation-maximization (EM) algorithm. A single diffusion process is constructed for the jamming, while the state of targets are estimated through direct posterior inference, enhancing computational efficiency. The noise variance is also estimated through EM algorithm. Numerical experiments demonstrate the effectiveness of the proposed method in structured jamming scenarios. The proposed DMDD algorithm achieves superior target detection performance, compared with existing methods.

eess.SP↗

arXiv2025

Cross-attention Secretly Performs Orthogonal Alignment in Recommendation Models

Cross-domain sequential recommendation (CDSR) aims to align heterogeneous user behavior sequences collected from different domains. While cross-attention is widely used to enhance alignment and improve recommendation performance, its underlying mechanism is not fully understood. Most researchers interpret cross-attention as residual alignment, where the output is generated by removing redundant and preserving non-redundant information from the query input by referencing another domain data which is input key and value. Beyond the prevailing view, we introduce Orthogonal Alignment, a phenomenon in which cross-attention discovers novel information that is not present in the query input, and further argue that those two contrasting alignment mechanisms can co-exist in recommendation models We find that when the query input and output of cross-attention are orthogonal, model performance improves over 300 experiments. Notably, Orthogonal Alignment emerges naturally, without any explicit orthogonality constraints. Our key insight is that Orthogonal Alignment emerges naturally because it improves scaling law. We show that baselines additionally incorporating cross-attention module outperform parameter-matched baselines, achieving a superior accuracy-per-model parameter. We hope these findings offer new directions for parameter-efficient scaling in multi-modal research.

cs.LG↗

arXiv2025

FastUMI-100K: Advancing Data-driven Robotic Manipulation with a Large-scale UMI-style Dataset

Data-driven robotic manipulation learning depends on large-scale, high-quality expert demonstration datasets. However, existing datasets, which primarily rely on human teleoperated robot collection, are limited in terms of scalability, trajectory smoothness, and applicability across different robotic embodiments in real-world environments. In this paper, we present FastUMI-100K, a large-scale UMI-style multimodal demonstration dataset, designed to overcome these limitations and meet the growing complexity of real-world manipulation tasks. Collected by FastUMI, a novel robotic system featuring a modular, hardware-decoupled mechanical design and an integrated lightweight tracking system, FastUMI-100K offers a more scalable, flexible, and adaptable solution to fulfill the diverse requirements of real-world robot demonstration data. Specifically, FastUMI-100K contains over 100K+ demonstration trajectories collected across representative household environments, covering 54 tasks and hundreds of object types. Our dataset integrates multimodal streams, including end-effector states, multi-view wrist-mounted fisheye images and textual annotations. Each trajectory has a length ranging from 120 to 500 frames. Experimental results demonstrate that FastUMI-100K enables high policy success rates across various baseline algorithms, confirming its robustness, adaptability, and real-world applicability for solving complex, dynamic manipulation challenges. The source code and dataset will be released in this link https://github.com/MrKeee/FastUMI-100K.

cs.RO↗

arXiv2025

AerialVG: A Challenging Benchmark for Aerial Visual Grounding by Exploring Positional Relations

Visual grounding (VG) aims to localize target objects in an image based on natural language descriptions. In this paper, we propose AerialVG, a new task focusing on visual grounding from aerial views. Compared to traditional VG, AerialVG poses new challenges, \emph{e.g.}, appearance-based grounding is insufficient to distinguish among multiple visually similar objects, and positional relations should be emphasized. Besides, existing VG models struggle when applied to aerial imagery, where high-resolution images cause significant difficulties. To address these challenges, we introduce the first AerialVG dataset, consisting of 5K real-world aerial images, 50K manually annotated descriptions, and 103K objects. Particularly, each annotation in AerialVG dataset contains multiple target objects annotated with relative spatial relations, requiring models to perform comprehensive spatial reasoning. Furthermore, we propose an innovative model especially for the AerialVG task, where a Hierarchical Cross-Attention is devised to focus on target regions, and a Relation-Aware Grounding module is designed to infer positional relations. Experimental results validate the effectiveness of our dataset and method, highlighting the importance of spatial reasoning in aerial visual grounding. The code and dataset will be released.

cs.CV↗

arXiv2025

Embodied Intelligence in Disassembly: Multimodal Perception Cross-validation and Continual Learning in Neuro-Symbolic TAMP

With the rapid development of the new energy vehicle industry, the efficient disassembly and recycling of power batteries have become a critical challenge for the circular economy. In current unstructured disassembly scenarios, the dynamic nature of the environment severely limits the robustness of robotic perception, posing a significant barrier to autonomous disassembly in industrial applications. This paper proposes a continual learning framework based on Neuro-Symbolic task and motion planning (TAMP) to enhance the adaptability of embodied intelligence systems in dynamic environments. Our approach integrates a multimodal perception cross-validation mechanism into a bidirectional reasoning flow: the forward working flow dynamically refines and optimizes action strategies, while the backward learning flow autonomously collects effective data from historical task executions to facilitate continual system learning, enabling self-optimization. Experimental results show that the proposed framework improves the task success rate in dynamic disassembly scenarios from 81.68% to 100%, while reducing the average number of perception misjudgments from 3.389 to 1.128. This research provides a new paradigm for enhancing the robustness and adaptability of embodied intelligence in complex industrial environments.

cs.RO↗

arXiv2025

Decoding RobKiNet: Insights into Efficient Training of Robotic Kinematics Informed Neural Network

In robots task and motion planning (TAMP), it is crucial to sample within the robot's configuration space to meet task-level global constraints and enhance the efficiency of subsequent motion planning. Due to the complexity of joint configuration sampling under multi-level constraints, traditional methods often lack efficiency. This paper introduces the principle of RobKiNet, a kinematics-informed neural network, for end-to-end sampling within the Continuous Feasible Set (CFS) under multiple constraints in configuration space, establishing its Optimization Expectation Model. Comparisons with traditional sampling and learning-based approaches reveal that RobKiNet's kinematic knowledge infusion enhances training efficiency by ensuring stable and accurate gradient optimization.Visualizations and quantitative analyses in a 2-DOF space validate its theoretical efficiency, while its application on a 9-DOF autonomous mobile manipulator robot(AMMR) demonstrates superior whole-body and decoupled control, excelling in battery disassembly tasks. RobKiNet outperforms deep reinforcement learning with a training speed 74.29 times faster and a sampling accuracy of up to 99.25%, achieving a 97.33% task completion rate in real-world scenarios.

cs.RO↗

arXiv2025

Exploring Spatial Representation to Enhance LLM Reasoning in Aerial Vision-Language Navigation

Aerial Vision-and-Language Navigation (VLN) is a novel task enabling Unmanned Aerial Vehicles (UAVs) to navigate in outdoor environments through natural language instructions and visual cues. However, it remains challenging due to the complex spatial relationships in aerial scenes.In this paper, we propose a training-free, zero-shot framework for aerial VLN tasks, where the large language model (LLM) is leveraged as the agent for action prediction. Specifically, we develop a novel Semantic-Topo-Metric Representation (STMR) to enhance the spatial reasoning capabilities of LLMs. This is achieved by extracting and projecting instruction-related semantic masks onto a top-down map, which presents spatial and topological information about surrounding landmarks and grows during the navigation process. At each step, a local map centered at the UAV is extracted from the growing top-down map, and transformed into a ma trix representation with distance metrics, serving as the text prompt to LLM for action prediction in response to the given instruction. Experiments conducted in real and simulation environments have proved the effectiveness and robustness of our method, achieving absolute success rate improvements of 26.8% and 5.8% over current state-of-the-art methods on simple and complex navigation tasks, respectively. The dataset and code will be released soon.

cs.RO↗

arXiv2025

DYNAMO: Dynamic Neutral Atom Multi-programming Optimizer Towards Quantum Operating Systems

As quantum computing advances towards practical applications, quantum operating systems become inevitable, where multi-programming -- the core functionality of operating systems -- enables concurrent execution of multiple quantum programs to enhance hardware utilization. However, most quantum compilation work focuses solely on single-circuit execution, severely limiting resource efficiency and hindering quantum operating system development. We propose Dynamic Neutral Atom Multi-programming Optimizer (DYNAMO), a method that realizes multi-programming on neutral atom quantum architectures through parallel compilation and intelligent resource allocation across multiple quantum processing units (QPUs). DYNAMO addresses two critical challenges: inefficient and difficult resource partitioning, and complex scheduling conflicts from concurrent program. Our method enables efficient spatial and temporal resource sharing while maintaining circuit correctness and hardware constraints. Experimental evaluation across circuits ranging from 12 to over 1200 gates demonstrates that DYNAMO achieves up to 14.39x compilation speedup while reducing execution stages by an average of 50.47%. Furthermore, DYNAMO successfully distributes workloads across multiple QPUs with balanced resource utilization. By enabling efficient multi-programming capabilities, DYNAMO establishes a critical foundation towards realizing practical quantum operating systems.

quant-ph↗

arXiv2025

Cross from Left to Right Brain: Adaptive Text Dreamer for Vision-and-Language Navigation

Vision-and-Language Navigation (VLN) requires the agent to navigate by following natural instructions under partial observability, making it difficult to align perception with language. Recent methods mitigate this by imagining future scenes, yet they rely on vision-based synthesis, leading to high computational cost and redundant details. To this end, we propose to adaptively imagine key environmental semantics via \textit{language} form, enabling a more reliable and efficient strategy. Specifically, we introduce a novel Adaptive Text Dreamer (ATD), a dual-branch self-guided imagination policy built upon a large language model (LLM). ATD is designed with a human-like left-right brain architecture, where the left brain focuses on logical integration, and the right brain is responsible for imaginative prediction of future scenes. To achieve this, we fine-tune only the Q-former within both brains to efficiently activate domain-specific knowledge in the LLM, enabling dynamic updates of logical reasoning and imagination during navigation. Furthermore, we introduce a cross-interaction mechanism to regularize the imagined outputs and inject them into a navigation expert module, allowing ATD to jointly exploit both the reasoning capacity of the LLM and the expertise of the navigation model. We conduct extensive experiments on the R2R benchmark, where ATD achieves state-of-the-art performance with fewer parameters. The code is \href{https://github.com/zhangpingrui/Adaptive-Text-Dreamer}{here}.

cs.CV↗

arXiv2025

Superatomic hydrogen: achieving effective aggregation of hydrogen atoms at pressures lower than that of metallic hydrogen

Metal hydrogen exhibiting electron delocalization properties has been recognized as an important prospect for achieving controlled nuclear fusion, but the extreme pressure conditions required exceeding hundreds of GPa remain a daunting challenge. Here, we propose a model of superatomic hydrogen, aiming to reduce the pressure conditions required for the effective aggregation of elemental hydrogen atoms. High-precision ab initio calculations indicate that the pressure required to compress the H13 system with one central atom and 12 surrounding atoms into a superatomic state is approximately two orders of magnitude lower than that of metallic hydrogen. Atomic-level analyses reveal that in the superatomic state of compressed H13, the central H atom donates its electron, and all electrons are delocalized on the superatomic molecular orbitals, which conforms to properties of metallic hydrogen. Our discovery in principle opens up the prospect of superatomic hydrogen in areas such as nuclear fusion.

physics.comp-ph↗

arXiv2025

Think Small, Act Big: Primitive Prompt Learning for Lifelong Robot Manipulation

Building a lifelong robot that can effectively leverage prior knowledge for continuous skill acquisition remains significantly challenging. Despite the success of experience replay and parameter-efficient methods in alleviating catastrophic forgetting problem, naively applying these methods causes a failure to leverage the shared primitives between skills. To tackle these issues, we propose Primitive Prompt Learning (PPL), to achieve lifelong robot manipulation via reusable and extensible primitives. Within our two stage learning scheme, we first learn a set of primitive prompts to represent shared primitives through multi-skills pre-training stage, where motion-aware prompts are learned to capture semantic and motion shared primitives across different skills. Secondly, when acquiring new skills in lifelong span, new prompts are appended and optimized with frozen pretrained prompts, boosting the learning via knowledge transfer from old skills to new ones. For evaluation, we construct a large-scale skill dataset and conduct extensive experiments in both simulation and real-world tasks, demonstrating PPL's superior performance over state-of-the-art methods.

cs.RO↗

arXiv2025

EvoMoE: Expert Evolution in Mixture of Experts for Multimodal Large Language Models

Recent advancements have shown that the Mixture of Experts (MoE) approach significantly enhances the capacity of large language models (LLMs) and improves performance on downstream tasks. Building on these promising results, multi-modal large language models (MLLMs) have increasingly adopted MoE techniques. However, existing multi-modal MoE tuning methods typically face two key challenges: expert uniformity and router rigidity. Expert uniformity occurs because MoE experts are often initialized by simply replicating the FFN parameters from LLMs, leading to homogenized expert functions and weakening the intended diversification of the MoE architecture. Meanwhile, router rigidity stems from the prevalent use of static linear routers for expert selection, which fail to distinguish between visual and textual tokens, resulting in similar expert distributions for image and text. To address these limitations, we propose EvoMoE, an innovative MoE tuning framework. EvoMoE introduces a meticulously designed expert initialization strategy that progressively evolves multiple robust experts from a single trainable expert, a process termed expert evolution that specifically targets severe expert homogenization. Furthermore, we introduce the Dynamic Token-aware Router (DTR), a novel routing mechanism that allocates input tokens to appropriate experts based on their modality and intrinsic token values. This dynamic routing is facilitated by hypernetworks, which dynamically generate routing weights tailored for each individual token. Extensive experiments demonstrate that EvoMoE significantly outperforms other sparse MLLMs across a variety of multi-modal benchmarks, including MME, MMBench, TextVQA, and POPE. Our results highlight the effectiveness of EvoMoE in enhancing the performance of MLLMs by addressing the critical issues of expert uniformity and router rigidity.

cs.CL↗

arXiv2025

SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Model

In this paper, we claim that spatial understanding is the keypoint in robot manipulation, and propose SpatialVLA to explore effective spatial representations for the robot foundation model. Specifically, we introduce Ego3D Position Encoding to inject 3D information into the input observations of the visual-language-action model, and propose Adaptive Action Grids to represent spatial robot movement actions with adaptive discretized action grids, facilitating learning generalizable and transferrable spatial action knowledge for cross-robot control. SpatialVLA is first pre-trained on top of a vision-language model with 1.1 Million real-world robot episodes, to learn a generalist manipulation policy across multiple robot environments and tasks. After pre-training, SpatialVLA is directly applied to perform numerous tasks in a zero-shot manner. The superior results in both simulation and real-world robots demonstrate its advantage of inferring complex robot motion trajectories and its strong in-domain multi-task generalization ability. We further show the proposed Adaptive Action Grids offer a new and effective way to fine-tune the pre-trained SpatialVLA model for new simulation and real-world setups, where the pre-learned action grids are re-discretized to capture robot-specific spatial action movements of new setups. The superior results from extensive evaluations demonstrate the exceptional in-distribution generalization and out-of-distribution adaptation capability, highlighting the crucial benefit of the proposed spatial-aware representations for generalist robot policy learning. All the details and codes will be open-sourced.

cs.RO↗