arXiv subjects

Zhuofan Zhang

Publications and source records attributed to Zhuofan Zhang.

14 recordsLinked to original sources

arXiv2026

EvoNav-Bench: Benchmarking Lifelong Navigation in Evolving Environments

Lifelong navigation (LN) requires an embodied agent to solve a sequence of navigation subtasks in the same environment. Since solving each subtask from scratch incurs redundant exploration, an LN agent must consolidate experience from earlier stages and reuse it in later stages, often through persistent scene representations such as scene graphs or visual snapshots. However, existing approaches typically assume a stationary environment, whereas in real-world LN settings, human activities can cause the environment to evolve. With the stationary assumption violated, existing methods may fuse outdated prior observations with new observations, yet current benchmarks cannot reveal this failure mode. In this paper, we present EvoNav-Bench, which extends the GOAT-Bench style LN formulation in the context of evolving environments. Built on the ProcTHOR framework, EvoNav-Bench introduces environment modifications between navigation tasks, making prior experience useful but not fully reliable. This design enables controlled evaluation of how environment evolution affects LN agents that reuse prior scene observations. Using EvoNav-Bench, we benchmark three recent methods that build and reuse scene representations for navigation. We also compare three simple heuristic strategies for handling environment evolution: Frontier-Update, Fail-then-Update, and Stage-Reset. Our results show that existing methods are brittle under environment evolution, while the heuristic strategies enable a controlled analysis of how agents can adapt to scene changes and mitigate their impact.

cs.RO↗

arXiv2026

Generation of dense relativistic electron beams via vortex laser-driven self-generated magnetic pinching

In multi-petawatt laser plasma accelerators, achieving high-density relativistic electron beams is typically accompanied by large transverse divergence, limiting the attainable effective electron density needed for high-flux interaction regimes relevant to laboratory astrophysics. Here we report experimental demonstration of self-generated magnetic pinching (SMP), a collective mechanism that actively regulates transverse beam dynamics using a Laguerre-Gaussian laser at strong relativistic intensity (~8 x 10^19 W/cm^2) interacting with an underdense plasma. The electron beam evolves from a two-lobe high-charge injection structure into a compressed, high-density profile, yielding a threefold reduction in divergence and nearly an order-of-magnitude enhancement in effective beam density compared with a Gaussian driver. Particle-in-cell simulations agree with the experimental observations and reveal that a self-generated azimuthal magnetic field governs the electron dynamics within the SMP regime, which is defined by the forming condition S = 0.717 l a0 [ne(10^18 cm^-3)]^-3/4 = 1, where l, a0, and ne are topological charge, laser amplitude, and plasma density, respectively. A transient kick from a dense inner sheath electron population drives collective magnetic pinching, transforming an initially separated electron distribution into a compressed and well-collimated beam. For higher-power laser systems, the forming condition can be extended to higher plasma densities and larger orbital angular momentum modes, potentially enabling electron beams with charges exceeding several nC and effective densities above 10^19 cm^-3. This mechanism provides a route to overcoming transverse expansion and enhancing rare interaction processes relevant to high-flux particle sources.

physics.plasm-ph↗

arXiv2026

UniLM-Nav: A Unified Framework for Zero-Shot Last-Mile Navigation

Mobile manipulation requires a robot to navigate to a target object or receptacle and then perform intended manipulation. However, reaching the vicinity of the target does not guarantee a manipulation-ready base pose, a problem known as last-mile navigation. Prior methods for last-mile navigation either rely on manual pose annotation or task-specific training, limiting their scalability to open-vocabulary settings with fine-grained spatial constraints. We propose UniLM-Nav, a unified framework for zero-shot open-vocabulary last-mile navigation. UniLM-Nav decomposes last-mile navigation into view selection, task-conditioned affordance grounding, and geometry-aware base-pose reasoning, all resolved with a shared multimodal large language model (MLLM) backend. Specifically, UniLM-Nav first selects a reference view that best captures the target object or receptacle from recently collected observations. It then grounds task-relevant affordance point in the selected view and lifts the result into the robot-centric coordinate frame. Finally, conditioned on the grounded affordance, task context, and robot geometry, it infers a manipulation-ready base pose for the robot. We evaluate UniLM-Nav on the OVMM benchmark, where it outperforms the previous state-of-the-art method, MoTo, by 3.13 percentage points. Analyses show that the components of our method are crucial to final performance, and that the choice of MLLM also has a substantial effect. We further deploy UniLM-Nav on a Unitree B2 quadruped robot with a 6-DoF Unitree Z1 manipulator, validating its applicability to real-world mobile manipulation tasks.

cs.RO↗

arXiv2026

Plasma-state metasurfaces for ultra-intensive field manipulation

High-power lasers offer ultrahigh intensities for plasma interactions, but they lack advanced techniques to control the properties of the fields, because no optical elements could withstand their high intensities. The vibrant field of metasurfaces has transformed modern optics by enabling unprecedented control over light at subwavelength through deliberate design. However, metasurfaces have traditionally been limited to solid-state materials and low light intensities. Extending the sophisticated capabilities of metasurfaces from solids into the plasma realm would open new horizons for high-field science. Here, we experimentally demonstrate plasma-state metasurfaces (PSMs) through the photonic spin Hall effect and stable-propagating vortex beam generation irradiated by intense light. Time-resolved pump-probe measurements reveal that the functionality of PSMs can persist for several picoseconds, making them suitable for controlling ultra-intense femtosecond lasers, even in state-of-the-art multi-petawatt systems. Harnessing the powerful toolkit of metasurfaces, this approach holds the promise to revolutionize our ability to manipulate the amplitude, phase, polarization, and wavefront of high-power lasers during their pulse duration. It also opens new possibilities for innovative applications in laser-plasma interactions such as compact particle acceleration and novel radiation sources.

physics.plasm-ph↗

arXiv2026

Probabilistic Abstract Interpretation on Neural Networks via Grids Approximation

Probabilistic abstract interpretation is a theory used to extract particular properties of a computer program when it is infeasible to test every single inputs. In this paper we apply the theory on neural networks for the same purpose: to analyse density distribution flow of all possible inputs of a neural network when a network has uncountably many or countable but infinitely many inputs. We show how this theoretical framework works in neural networks and then discuss different abstract domains and corresponding Moore-Penrose pseudo-inverses together with abstract transformers used in the framework. We also present experimental examples to show how this framework helps to analyse real world problems.

cs.AI↗

arXiv2026

Distribution and Clusters Approximations as Abstract Domains in Probabilistic Abstract Interpretation to Neural Network Analysis

The probabilistic abstract interpretation framework of neural network analysis analyzes a neural network by analyzing its density distribution flow of all possible inputs. The grids approximation is one of abstract domains the framework uses which abstracts concrete space into grids. In this paper, we introduce two novel approximation methods: distribution approximation and clusters approximation. We show how these two methods work in theory with corresponding abstract transformers with help of illustrations of some simple examples.

cs.AI↗

arXiv2025

Neural network-based deconvolution for GeV-Scale Gamma-Ray Spectroscopy

High-energy gamma-ray spectroscopy is crucial for studying and advancing the application of high-energy photons in areas like strong-field physics, high-energy-density science, and laboratory astrophysics. However, high-energy gamma-ray spectroscopy in the multi-MeV to GeV range faces significant challenges in precise spectral reconstruction. This study presents a machine learning-based inversion approach that combines a spectrometer design with advanced deconvolution algorithms. We develop a gamma-ray spectrometer optimized through Monte Carlo simulations for maximum positron yield and minimal noise. A two-stage neural network framework is proposed based on the structure of the spectrometer: a denoising autoencoder suppresses statistical noise in measured positron spectra, while a U-Net architecture solves the ill-posed inverse problem to reconstruct incident gamma spectra. This approach establishes a new methodology for gamma-ray diagnostics in strong-field QED experiments and high-energy photon sources.

physics.ins-det↗

arXiv2025

From Objects to Anywhere: A Holistic Benchmark for Multi-level Visual Grounding in 3D Scenes

3D visual grounding has made notable progress in localizing objects within complex 3D scenes. However, grounding referring expressions beyond objects in 3D scenes remains unexplored. In this paper, we introduce Anywhere3D-Bench, a holistic 3D visual grounding benchmark consisting of 2,886 referring expression-3D bounding box pairs spanning four different grounding levels: human-activity areas, unoccupied space beyond objects, individual objects in the scene, and fine-grained object parts. We assess a range of state-of-the-art 3D visual grounding methods alongside large language models (LLMs) and multimodal LLMs (MLLMs) on Anywhere3D-Bench. Experimental results reveal that space-level and part-level visual grounding pose the greatest challenges: space-level tasks require a more comprehensive spatial reasoning ability, for example, modeling distances and spatial relations within 3D space, while part-level tasks demand fine-grained perception of object composition. Even the best-performing models, Google Gemini-2.5-Pro and OpenAI o3, achieve just around 30% accuracy on space-level tasks and around 40% on part-level tasks, significantly lower than its performance on area-level and object-level tasks. These findings underscore a critical gap in current models' capacity to understand and reason about 3D scenes beyond object-level semantics.

cs.CV↗

arXiv2025

Move to Understand a 3D Scene: Bridging Visual Grounding and Exploration for Efficient and Versatile Embodied Navigation

Embodied scene understanding requires not only comprehending visual-spatial information that has been observed but also determining where to explore next in the 3D physical world. Existing 3D Vision-Language (3D-VL) models primarily focus on grounding objects in static observations from 3D reconstruction, such as meshes and point clouds, but lack the ability to actively perceive and explore their environment. To address this limitation, we introduce \underline{\textbf{M}}ove \underline{\textbf{t}}o \underline{\textbf{U}}nderstand (\textbf{\model}), a unified framework that integrates active perception with \underline{\textbf{3D}} vision-language learning, enabling embodied agents to effectively explore and understand their environment. This is achieved by three key innovations: 1) Online query-based representation learning, enabling direct spatial memory construction from RGB-D frames, eliminating the need for explicit 3D reconstruction. 2) A unified objective for grounding and exploring, which represents unexplored locations as frontier queries and jointly optimizes object grounding and frontier selection. 3) End-to-end trajectory learning that combines \textbf{V}ision-\textbf{L}anguage-\textbf{E}xploration pre-training over a million diverse trajectories collected from both simulated and real-world RGB-D sequences. Extensive evaluations across various embodied navigation and question-answering benchmarks show that MTU3D outperforms state-of-the-art reinforcement learning and modular navigation approaches by 14\%, 23\%, 9\%, and 2\% in success rate on HM3D-OVON, GOAT-Bench, SG3D, and A-EQA, respectively. \model's versatility enables navigation using diverse input modalities, including categories, language descriptions, and reference images. These findings highlight the importance of bridging visual grounding and exploration for embodied intelligence.

cs.CV↗

arXiv2025

Task-oriented Sequential Grounding and Navigation in 3D Scenes

Grounding natural language in 3D environments is a critical step toward achieving robust 3D vision-language alignment. Current datasets and models for 3D visual grounding predominantly focus on identifying and localizing objects from static, object-centric descriptions. These approaches do not adequately address the dynamic and sequential nature of task-oriented scenarios. In this work, we introduce a novel task: Task-oriented Sequential Grounding and Navigation in 3D Scenes, where models must interpret step-by-step instructions for daily activities by either localizing a sequence of target objects in indoor scenes or navigating toward them within a 3D simulator. To facilitate this task, we present SG3D, a large-scale dataset comprising 22,346 tasks with 112,236 steps across 4,895 real-world 3D scenes. The dataset is constructed by combining RGB-D scans from various 3D scene datasets with an automated task generation pipeline, followed by human verification for quality assurance. We benchmark contemporary methods on SG3D, revealing the significant challenges in understanding task-oriented context across multiple steps. Furthermore, we propose SG-LLM, a state-of-the-art approach leveraging a stepwise grounding paradigm to tackle the sequential grounding task. Our findings underscore the need for further research to advance the development of more capable and context-aware embodied agents.

cs.CV↗

arXiv2024

Rethinking Structure Learning For Graph Neural Networks

To improve the performance of Graph Neural Networks (GNNs), Graph Structure Learning (GSL) has been extensively applied to reconstruct or refine original graph structures, effectively addressing issues like heterophily, over-squashing, and noisy structures. While GSL is generally thought to improve GNN performance, it often leads to longer training times and more hyperparameter tuning. Besides, the distinctions among current GSL methods remain ambiguous from the perspective of GNN training, and there is a lack of theoretical analysis to quantify their effectiveness. Recent studies further suggest that, under fair comparisons with the same hyperparameter tuning, GSL does not consistently outperform baseline GNNs. This motivates us to ask a critical question: is GSL really useful for GNNs? To address this question, this paper makes two key contributions. First, we propose a new GSL framework, which includes three steps: GSL base (the representation used for GSL) construction, new structure construction, and view fusion, to better understand the effectiveness of GSL in GNNs. Second, after graph convolution, we analyze the differences in mutual information (MI) between node representations derived from the original topology and those from the newly constructed topology. Surprisingly, our empirical observations and theoretical analysis show that no matter which type of graph structure construction methods are used, after feeding the same GSL bases to the newly constructed graph, there is no MI gain compared to the original GSL bases. To fairly reassess the effectiveness of GSL, we conduct ablation experiments and find that it is the pretrained GSL bases that enhance GNN performance, and in most cases, GSL cannot improve GNN performance. This finding encourages us to rethink the essential components in GNNs, such as self-training and structural encoding, in GNN design rather than GSL.

cs.LG↗

arXiv2024

Unifying 3D Vision-Language Understanding via Promptable Queries

A unified model for 3D vision-language (3D-VL) understanding is expected to take various scene representations and perform a wide range of tasks in a 3D scene. However, a considerable gap exists between existing methods and such a unified model, due to the independent application of representation and insufficient exploration of 3D multi-task training. In this paper, we introduce PQ3D, a unified model capable of using Promptable Queries to tackle a wide range of 3D-VL tasks, from low-level instance segmentation to high-level reasoning and planning. This is achieved through three key innovations: (1) unifying various 3D scene representations (i.e., voxels, point clouds, multi-view images) into a shared 3D coordinate space by segment-level grouping, (2) an attention-based query decoder for task-specific information retrieval guided by prompts, and (3) universal output heads for different tasks to support multi-task training. Tested across ten diverse 3D-VL datasets, PQ3D demonstrates impressive performance on these tasks, setting new records on most benchmarks. Particularly, PQ3D improves the state-of-the-art on ScanNet200 by 4.9% (AP25), ScanRefer by 5.4% (acc@0.5), Multi3DRefer by 11.7% (F1@0.5), and Scan2Cap by 13.4% (CIDEr@0.5). Moreover, PQ3D supports flexible inference with individual or combined forms of available 3D representations, e.g., solely voxel input.

cs.CV↗

arXiv2023

Minimum Latency Deep Online Video Stabilization

We present a novel camera path optimization framework for the task of online video stabilization. Typically, a stabilization pipeline consists of three steps: motion estimating, path smoothing, and novel view rendering. Most previous methods concentrate on motion estimation, proposing various global or local motion models. In contrast, path optimization receives relatively less attention, especially in the important online setting, where no future frames are available. In this work, we adopt recent off-the-shelf high-quality deep motion models for motion estimation to recover the camera trajectory and focus on the latter two steps. Our network takes a short 2D camera path in a sliding window as input and outputs the stabilizing warp field of the last frame in the window, which warps the coming frame to its stabilized position. A hybrid loss is well-defined to constrain the spatial and temporal consistency. In addition, we build a motion dataset that contains stable and unstable motion pairs for the training. Extensive experiments demonstrate that our approach significantly outperforms state-of-the-art online methods both qualitatively and quantitatively and achieves comparable performance to offline methods. Our code and dataset are available at https://github.com/liuzhen03/NNDVS

cs.CV↗

arXiv2021

The Effect of Training Parameters and Mechanisms on Decentralized Federated Learning based on MNIST Dataset

Federated Learning is an algorithm suited for training models on decentralized data, but the requirement of a central "server" node is a bottleneck. In this document, we first introduce the notion of Decentralized Federated Learning (DFL). We then perform various experiments on different setups, such as changing model aggregation frequency, switching from independent and identically distributed (IID) dataset partitioning to non-IID partitioning with partial global sharing, using different optimization methods across clients, and breaking models into segments with partial sharing. All experiments are run on the MNIST handwritten digits dataset. We observe that those altered training procedures are generally robust, albeit non-optimal. We also observe failures in training when the variance between model weights is too large. The open-source experiment code is accessible through GitHub\footnote{Code was uploaded at \url{https://github.com/zhzhang2018/DecentralizedFL}}.

cs.LG↗