arXiv subjects

Dong Wang

Publications and source records attributed to Dong Wang.

At least 181 records · Page 10Linked to original sources

arXiv2025

In Pursuit of Pixel Supervision for Visual Pre-training

At the most basic level, pixels are the source of the visual information through which we perceive the world. Pixels contain information at all levels, ranging from low-level attributes to high-level concepts. Autoencoders represent a classical and long-standing paradigm for learning representations from pixels or other raw inputs. In this work, we demonstrate that autoencoder-based self-supervised learning remains competitive today and can produce strong representations for downstream tasks, while remaining simple, stable, and efficient. Our model, codenamed "Pixio", is an enhanced masked autoencoder (MAE) with more challenging pre-training tasks and more capable architectures. The model is trained on 2B web-crawled images with a self-curation strategy with minimal human curation. Pixio performs competitively across a wide range of downstream tasks in the wild, including monocular depth estimation (e.g., Depth Anything), feed-forward 3D reconstruction (i.e., MapAnything), semantic segmentation, and robot learning, outperforming or matching DINOv3 trained at similar scales. Our results suggest that pixel-space self-supervised learning can serve as a promising alternative and a complement to latent-space approaches.

cs.CV↗

arXiv2025

TCLeaf-Net: a transformer-convolution framework with global-local attention for robust in-field lesion-level plant leaf disease detection

Timely and accurate detection of foliar diseases is vital for safeguarding crop growth and reducing yield losses. Yet, in real-field conditions, cluttered backgrounds, domain shifts, and limited lesion-level datasets hinder robust modeling. To address these challenges, we release Daylily-Leaf, a paired lesion-level dataset comprising 1,746 RGB images and 7,839 lesions captured under both ideal and in-field conditions, and propose TCLeaf-Net, a transformer-convolution hybrid detector optimized for real-field use. TCLeaf-Net is designed to tackle three major challenges. To mitigate interference from complex backgrounds, the transformer-convolution module (TCM) couples global context with locality-preserving convolution to suppress non-leaf regions. To reduce information loss during downsampling, the raw-scale feature recalling and sampling (RSFRS) block combines bilinear resampling and convolution to preserve fine spatial detail. To handle variations in lesion scale and feature shifts, the deformable alignment block with FPN (DFPN) employs offset-based alignment and multi-receptive-field perception to strengthen multi-scale fusion. Experimental results show that on the in-field split of the Daylily-Leaf dataset, TCLeaf-Net improves mAP@50 by 5.4 percentage points over the baseline model, reaching 78.2\%, while reducing computation by 7.5 GFLOPs and GPU memory usage by 8.7\%. Moreover, the model outperforms recent YOLO and RT-DETR series in both precision and recall, and demonstrates strong performance on the PlantDoc, Tomato-Leaf, and Rice-Leaf datasets, validating its robustness and generalizability to other plant disease detection scenarios.

cs.CV↗

arXiv2025

Equilibrium Policy Generalization: A Reinforcement Learning Framework for Cross-Graph Zero-Shot Generalization in Pursuit-Evasion Games

Equilibrium learning in adversarial games is an important topic widely examined in the fields of game theory and reinforcement learning (RL). Pursuit-evasion game (PEG), as an important class of real-world games from the fields of robotics and security, requires exponential time to be accurately solved. When the underlying graph structure varies, even the state-of-the-art RL methods require recomputation or at least fine-tuning, which can be time-consuming and impair real-time applicability. This paper proposes an Equilibrium Policy Generalization (EPG) framework to effectively learn a generalized policy with robust cross-graph zero-shot performance. In the context of PEGs, our framework is generally applicable to both pursuer and evader sides in both no-exit and multi-exit scenarios. These two generalizability properties, to our knowledge, are the first to appear in this domain. The core idea of the EPG framework is to train an RL policy across different graph structures against the equilibrium policy for each single graph. To construct an equilibrium oracle for single-graph policies, we present a dynamic programming (DP) algorithm that provably generates pure-strategy Nash equilibrium with near-optimal time complexity. To guarantee scalability with respect to pursuer number, we further extend DP and RL by designing a grouping mechanism and a sequence model for joint policy decomposition, respectively. Experimental results show that, using equilibrium guidance and a distance feature proposed for cross-graph PEG training, the EPG framework guarantees desirable zero-shot performance in various unseen real-world graphs. Besides, when trained under an equilibrium heuristic proposed for the graphs with exits, our generalized pursuer policy can even match the performance of the fine-tuned policies from the state-of-the-art PEG methods.

cs.LG↗

arXiv2025

Primal-dual policy learning for mean-field stochastic LQR problem

Integrating data-driven techniques with mechanism-driven insights has recently gained popularity as a powerful learning approach to solving traditional LQR problems for designing intelligent controllers in complex dynamic systems. However, the theoretical understanding of various reinforcement learning algorithms needs further exploration to enhance their efficiency and safety. In this article, by means of primal-dual optimization tools, we study the partially model-free design of the mean-field stochastic LQR (MF-SLQR) controller using a policy learning approach. Firstly, by designing appropriate optimizing variables, the considered MF-SLQR problem is transformed into a new static nonconvex constrained optimization problem with equivalence preserved in certain senses. After that, the equivalent formulation of the duality results is constructed via finding the solution of the generalized Lyapunov equation. Then, the strong duality is analyzed, based on which we establish a primal-dual algorithm by Karush-Kuhn-Tucker conditions. More importantly, a partially model-free implementation is also presented, which has a direct connection with the classical policy iteration algorithm. Finally, we use a high-dimensional example to validate our methods.

math.OC↗

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

UMind-VL: A Generalist Ultrasound Vision-Language Model for Unified Grounded Perception and Comprehensive Interpretation

Despite significant strides in medical foundation models, the ultrasound domain lacks a comprehensive solution capable of bridging low-level Ultrasound Grounded Perception (e.g., segmentation, localization) and high-level Ultrasound Comprehensive Interpretation (e.g., diagnosis, reasoning). To bridge this gap, we propose UMind-VL, a unified foundation model designed to synergize pixel-level structural understanding with complex clinical reasoning. We first introduce UMind-DS, a large-scale multimodal dataset comprising 1.2 million ultrasound image-text pairs across 16 anatomical regions, enriching standard data with pixel-level annotations and clinician-validated rationales. Architecturally, UMind-VL incorporates a lightweight Dynamic Convolutional Mask Decoder that generates masks via dynamic kernels conditioned on LLM outputs. This design, combined with task-specific tokens, unifies segmentation, detection, geometric measurement, and diagnosis tasks within a single framework. Extensive evaluations demonstrate that UMind-VL significantly outperforms existing generalist multimodal models and achieves performance on par with, or superior to, state-of-the-art specialist models across segmentation, detection, keypoint localization, and diagnostic reasoning benchmarks, while maintaining strong generalization ability. We demonstrate the capability of UMind-VL in Figure 1.

cs.CV↗

arXiv2025

FunReason: Enhancing Large Language Models' Function Calling via Self-Refinement Multiscale Loss and Automated Data Refinement

The integration of large language models (LLMs) with function calling has emerged as a crucial capability for enhancing their practical utility in real-world applications. However, effectively combining reasoning processes with accurate function execution remains a significant challenge. Traditional training approaches often struggle to balance the detailed reasoning steps with the precision of function calls, leading to suboptimal performance. To address these limitations, we introduce FunReason, a novel framework that enhances LLMs' function calling capabilities through an automated data refinement strategy and a Self-Refinement Multiscale Loss (SRML) approach. FunReason leverages LLMs' natural reasoning abilities to generate high-quality training examples, focusing on query parseability, reasoning coherence, and function call precision. The SRML approach dynamically balances the contribution of reasoning processes and function call accuracy during training, addressing the inherent trade-off between these two critical aspects. FunReason achieves performance comparable to GPT-4o while effectively mitigating catastrophic forgetting during fine-tuning. FunReason provides a comprehensive solution for enhancing LLMs' function calling capabilities by introducing a balanced training methodology and a data refinement pipeline. For code and dataset, please refer to our repository at GitHub https://github.com/BingguangHao/FunReason

cs.LG↗

arXiv2025

Topmetal-L: A Low Noise Charge-Sensitive Pixel Sensor for POLAR2/LPD

POLAR-2 is a next-generation space astronomy platform led by China, with its core scientific objective focused on high-precision polarization measurements of gamma-ray bursts. As one of its key payloads, the Low-energy Polarization Detector (LPD) is designed to perform wide-field surveys to capture X-ray polarization information from gamma-ray bursts in the 2$\sim$10 keV energy range. This paper presents Topmetal-L, a dedicated charge-sensitive pixel sensor developed for the LPD prototype upgrade. Fabricated in a 130 nm CMOS process in 2024, the chip integrates a 356 $\times$ 512 pixel array with a pixel pitch of 45 $μ$m. Each pixel incorporates a 26 $\times$ 26 $μ$m$^2$ charge-collecting window and is capable of simultaneously outputting both energy and position information of deposited charges. Topmetal-L has been systematically optimized for power consumption, noise performance, and readout efficiency. It exhibits an input dynamic range of 0$\sim$4 ke$^{-}$, a typical charge-to-voltage conversion gain of 76.04 $μ$V/e$^{-}$, an average equivalent noise charge of approximately 22.8 e$^{-}$, a sensitive area exceeding 3.69 cm$^2$, and a total power consumption of 720 mW per chip. To meet the requirements of large-area, high-frame-rate readout for gas-based polarization detectors, a sentinel readout scheme is proposed, reducing the full-frame readout time to 730 $μ$s. A prototype Topmetal-L-based gas polarization detection system was evaluated across key energies: it exhibited a residual modulation of 0.26% $\pm$ 0.45% at 5.90 keV, a modulation factor of 66.67% $\pm$ 0.45% for a linearly polarized 8.05 keV source, and a count rate saturated at 15 k counts$\cdot$cm$^{-2}$$\cdot$s$^{-1}$ when tested at 5.40 keV.

astro-ph.IM↗

arXiv2025

CADTrack: Learning Contextual Aggregation with Deformable Alignment for Robust RGBT Tracking

RGB-Thermal (RGBT) tracking aims to exploit visible and thermal infrared modalities for robust all-weather object tracking. However, existing RGBT trackers struggle to resolve modality discrepancies, which poses great challenges for robust feature representation. This limitation hinders effective cross-modal information propagation and fusion, which significantly reduces the tracking accuracy. To address this limitation, we propose a novel Contextual Aggregation with Deformable Alignment framework called CADTrack for RGBT Tracking. To be specific, we first deploy the Mamba-based Feature Interaction (MFI) that establishes efficient feature interaction via state space models. This interaction module can operate with linear complexity, reducing computational cost and improving feature discrimination. Then, we propose the Contextual Aggregation Module (CAM) that dynamically activates backbone layers through sparse gating based on the Mixture-of-Experts (MoE). This module can encode complementary contextual information from cross-layer features. Finally, we propose the Deformable Alignment Module (DAM) to integrate deformable sampling and temporal propagation, mitigating spatial misalignment and localization drift. With the above components, our CADTrack achieves robust and accurate tracking in complex scenarios. Extensive experiments on five RGBT tracking benchmarks verify the effectiveness of our proposed method. The source code is released at https://github.com/IdolLab/CADTrack.

cs.CV↗

arXiv2025

Deep learning-based computational method for soft matter dynamics: Deep Onsager-Machlup method

A deep learning-based computational method is proposed for soft matter dynamics -- the deep Onsager-Machlup method (DOMM). It combines the brute forces of deep neural networks (DNNs) with the fundamental physics principle -- Onsager-Machlup variational principle (OMVP). In the DOMM, the trial solution to the dynamics is constructed by DNNs that allow us to explore a rich and complex set of admissible functions. It outperforms the Ritz-type variational method where one has to impose carefully-chosen trial functions. This capability endows the DOMM with the potential to solve rather complex problems in soft matter dynamics that involve multiple physics with multiple slow variables, multiple scales, and multiple dissipative processes. Actually, the DOMM can be regarded as an extension of the deep Ritz method (DRM) developed by E and Yu that uses DNNs to solve static problems in physics. In this work, as the first step, we focus on the validation of the DOMM as a useful computational method by using it to solve several typical soft matter dynamic problems: particle diffusion in dilute solutions, and two-phase dynamics with and without hydrodynamics. The predicted results agree very well with the analytical solution or numerical solution from traditional computational methods. These results show the accuracy and convergence of DOMM and justify it as an alternative computational method for solving soft matter dynamics.

cond-mat.soft↗

arXiv2025

A Novel Pixel-Chip-Based Region-of-Interest Readout Circuit Design

This paper presents a novel pixel chip readout scheme: the Region-of-Interest Readout Circuit (ROIRC), which is designed for large area, large array pixel chips and Gas Pixel Detector (GPD). This design employs a sentinel pixel detection strategy, enabling rapid identification and prioritized readout of the pixel regions containing signal events. During the scanning readout of these signal events, ROIRC employs a Block-based readout approach, effectively minimizing the readout of non-signal pixels. The functionality of ROIRC has been successfully implemented on both the ASIC and FPGA platforms. In the tests of the ROIRC, the pixel chip embedded in the GPD is capable of detecting low-energy X-rays in the range of 2-10 keV and supports multiple event readouts, and the pixel chip can read out photo-electron signal events with the count rate up to 15k / (cm2 x s).

physics.ins-det↗

arXiv2025

Wide-Field X-ray Polarimetry for High Energy Astronomical Transients: First results of the pathfinder CXPD Cubesat Mission

The Low Energy Polarization Detector (LPD) is a key component of the next-generation large-scale Gamma-Ray Burst polarimeter, POLAR-2. It is designed for polarization observations of transient sources in the soft X-ray energy range with a wide field of view (FOV). To validate the key technologies required for wide-FOV X-ray polarization measurements, the Cosmic X-ray Polarization Detector (CXPD) CubeSat was developed as a prototype for the LPD. The CXPD is equipped with two Gas Microchannel Plate Pixel Detectors (GMPDs) that measure X-ray polarization via the photoelectric effect, where ejected photoelectrons produce ionization tracks in the gas which are imaged to reconstruct their emission directions. Laboratory calibrations of the modulation factor and energy spectra were successfully performed using linear polarized X-ray sources at 2.98 keV, 4.51 keV, 6.40 keV, and 8.05 keV. Since its launch in June 2023, the CXPD has successfully completed critical in-orbit technology verification. It has also performed polarization observations of two bright X-ray sources Sco X-1 and the transient Swift J1727.8-1613 yielding constraints on their polarization degrees and angles. Notably, this was the first time that an anti-coincidence detector had been implemented in an X-ray polarimeter, enabling in-orbit verification of the charged-particle background rejection algorithm. These results demonstrate the feasibility of wide-field soft X-ray polarization measurements and provide essential guidance for the development of the LPD for the POLAR-2 mission, thereby advancing the frontier of X-ray polarization astronomy.

astro-ph.IM↗

arXiv2025

FunReason-MT Technical Report: Advanced Data Synthesis Solution for Real-world Multi-Turn Tool-use

Function calling (FC) empowers large language models (LLMs) and autonomous agents to interface with external tools, a critical capability for solving complex, real-world problems. As this ability becomes increasingly central to advanced AI systems, the need for high-quality, multi-turn training data to develop and refine it cannot be overstated. Existing data synthesis methods, such as random environment sampling or multi-agent role-playing, are not powerful enough to generate high-quality data in real-world environments. Practical challenges come in three folds: targeted data synthesis, hard query construction, and multi-turn logical dependency. To address these structural deficiencies, we present FunReason-MT, a novel data synthesis framework for real-world multi-turn tool use. FunReason-MT resolves the complexity barrier in multi-turn FC data by employing 1) Environment-API Graph Interactions to gather varied high-quality trajectories with targeted tool, 2) Advanced Tool-Query Synthesis to simplify hard query construction, and 3) Guided Iterative Chain for sophisticated CoT generation. Evaluations on Berkeley Function-Calling Leaderboard (BFCLv3) demonstrate the power of our framework: a 4B model built upon FunReason-MT generated data achieves state-of-the-art performance among comparable-sized models. Further performance improvements on BFCLv4 confirm that FunReason-MT provides a reliable and robust source for agentic learning.

cs.AI↗

arXiv2025

An Efficient Unconditionally Energy-Stable Numerical Scheme for Bose--Einstein Condensate

A numerical framework is proposed and analyzed for computing the ground state of Bose--Einstein condensates. A gradient flow approach is developed, incorporating both a Lagrange multiplier to enforce the $L^2$ conservation and a free energy dissipation. An explicit approximation is applied to the chemical potential, combined with an exponential time differencing (ETD) operator to the diffusion part, as well a stabilizing operator, to obtain an intermediate numerical profile. Afterward, an $L^2$ normalization is applied at the next numerical stage. A theoretical analysis reveals a free energy dissipation under a maximum norm bound assumption for the numerical solution, and such a maximum norm bound could be recovered by a careful convergence analysis and error estimate. In the authors' knowledge, the proposed method is the first numerical work that preserves the following combined theoretical properties: (1) an explicit computation at each time step, (2) unconditional free energy dissipation, (3) $L^2$ norm conservation at each time step, (4) a theoretical justification of convergence analysis and optimal rate error estimate. Comprehensive numerical experiments validate these theoretical results, demonstrating excellent agreement with established reference solutions.

math.NA↗

arXiv2025

Lane Graph Extraction from Aerial Imagery via Lane Segmentation Refinement with Diffusion Models

The lane graph is critical for applications such as autonomous driving and lane-level route planning. While previous research has focused on extracting lane-level graphs from aerial imagery using convolutional neural networks (CNNs) followed by post-processing segmentation-to-graph algorithms, these methods often face challenges in producing sharp and complete segmentation masks. Challenges such as occlusions, variations in lighting, and changes in road texture can lead to incomplete and inaccurate lane masks, resulting in poor-quality lane graphs. To address these challenges, we propose a novel approach that refines the lane masks, output by a CNN, using diffusion models. Experimental results on a publicly available dataset demonstrate that our method outperforms existing methods based solely on CNNs or diffusion models, particularly in terms of graph connectivity. Our lane mask refinement approach enhances the quality of the extracted lane graph, yielding gains of approximately 1.5\% in GEO F1 and 3.5\% in TOPO F1 scores over the best-performing CNN-based method, and improvements of 28\% and 34\%, respectively, compared to a prior diffusion-based approach. Both GEO F1 and TOPO F1 scores are critical metrics for evaluating lane graph quality. Additionally, ablation studies are conducted to evaluate the individual components of our approach, providing insights into their respective contributions and effectiveness.

cs.CV↗

arXiv2025

MT-HuBERT: Self-Supervised Mix-Training for Few-Shot Keyword Spotting in Mixed Speech

Few-shot keyword spotting aims to detect previously unseen keywords with very limited labeled samples. A pre-training and adaptation paradigm is typically adopted for this task. While effective in clean conditions, most existing approaches struggle with mixed keyword spotting--detecting multiple overlapping keywords within a single utterance--a capability essential for real-world applications. We have previously proposed a pre-training approach based on Mix-Training (MT) to tackle the mixed keyword detection problem and demonstrated its efficiency. However, this approach is fully supervised, unable to utilize vast unlabeled data. To this end, we propose Mix-Training HuBERT (MT-HuBERT), a self-supervised learning (SSL) pre-training framework that implements the MT criterion during pre-training. MT-HuBERT predicts, in a self-supervised manner, the clean acoustic units of each constituent signal from contextual cues, in contrast to predicting compositional patterns of mixed speech. Experiments conducted on the Google Speech Commands (GSC v2) corpus demonstrate that our proposed MT-HuBERT consistently outperforms several state-of-the-art baselines in few-shot KWS tasks under both mixed and clean conditions.

cs.SD↗

arXiv2025

NaturalReasoning: Reasoning in the Wild with 2.8M Challenging Questions

Scaling reasoning capabilities beyond traditional domains such as math and coding is hindered by the lack of diverse and high-quality questions. To overcome this limitation, we introduce a scalable approach for generating diverse and challenging reasoning questions, accompanied by reference answers. We present NaturalReasoning, a comprehensive dataset comprising 2.8 million questions that span multiple domains, including STEM fields (e.g., Physics, Computer Science), Economics, Social Sciences, and more. We demonstrate the utility of the questions in NaturalReasoning through knowledge distillation experiments which show that NaturalReasoning can effectively elicit and transfer reasoning capabilities from a strong teacher model. Furthermore, we demonstrate that NaturalReasoning is also effective for unsupervised self-training using external reward models or self-rewarding. To foster future work, we publicly release NaturalReasoning at https://huggingface.co/datasets/facebook/natural_reasoning.

cs.CL↗

arXiv2025

A Big Step Forward? A User-Centric Examination of iOS App Privacy Report and Enhancements

The prevalent engagement with mobile apps underscores the importance of understanding their data practices. Transparency plays a crucial role in this context, ensuring users to be informed and give consent before any data access occurs. Apple introduced a new feature since iOS 15.2, App Privacy Report, to inform users about detailed insights into apps' data access and sharing. This feature continues Apple's trend of privacy-focused innovations (following Privacy Nutrition Labels), and has been marketed as a big step forward in user privacy. However, its real-world impacts on user privacy and control remain unexamined. We thus proposed an end-to-end study involving systematic assessment of the App Privacy Report's real-world benefits and limitations, LLM-enabled and multi-technique synthesized enhancements, and comprehensive evaluation from both system and user perspectives. Through a structured focus group study with twelve everyday iOS users, we explored their experiences, understanding, and perceptions of the feature, suggesting its limited practical impact resulting from missing important details. We identified two primary user concerns: the clarity of data access purpose and domain description. In response, we proposed enhancements including a purpose inference framework and domain clarification pipeline. We demonstrated the effectiveness and benefits of such enhancements for mobile app users. This work provides practical insights that could help enhance user privacy transparency and discusses areas for future research.

cs.SE↗