arXiv subjects

Juan Zhou

Publications and source records attributed to Juan Zhou.

15 recordsLinked to original sources

arXiv2026

Mamoda2.5: Enhancing Unified Multimodal Model with DiT-MoE

We present Mamoda2.5, a unified AR-Diffusion framework that seamlessly integrates multimodal understanding and generation within a single architecture. To efficiently enhance the model's generation capability, we equip the Diffusion Transformer backbone with a fine-grained Mixture-of-Experts (MoE) design (128 experts, Top-8 routing), yielding a 25B-parameter model that activates only 3B parameters, significantly reducing training costs while scaling up the model capacity. Mamoda2.5 achieves top-tier generation performance on VBench 2.0 and sets a new record in video editing quality, surpassing evaluated open-source models and matching the performance of current top-tier proprietary models, including the Kling O1 on OpenVE-Bench. Furthermore, we introduce a joint few-step distillation and reinforcement learning framework that compresses the 30-step editing model into a 4-step model and greatly accelerates model inference. Compared to open-source baselines, Mamoda2.5 achieves up to $95.9\times$ faster video editing inference. In real-world applications, Mamoda2.5 has been successfully deployed for content moderation and creative restoration tasks in advertising scenarios, achieving a 98% success rate in internal advertising video editing scenario.

cs.CV↗

arXiv2026

MSLAU-Net: A Hybrid CNN-Transformer Network for Medical Image Segmentation

Accurate medical image segmentation allows for the precise delineation of anatomical structures and pathological regions, which is essential for treatment planning, surgical navigation, and disease monitoring. Both CNN-based and Transformer-based methods have achieved remarkable success in medical image segmentation tasks. However, CNN-based methods struggle to effectively capture global contextual information due to the inherent limitations of convolution operations. Meanwhile, Transformer-based methods suffer from insufficient local feature modeling and face challenges related to the high computational complexity caused by the self-attention mechanism. To address these limitations, we propose a novel hybrid CNN-Transformer architecture, named MSLAU-Net, which integrates the strengths of both paradigms. The proposed MSLAU-Net incorporates two key ideas. First, it introduces Multi-Scale Linear Attention, designed to efficiently extract multi-scale features from medical images while modeling long-range dependencies with low computational complexity. Second, it adopts a top-down feature aggregation mechanism, which performs multi-level feature aggregation and restores spatial resolution using a lightweight structure. Extensive experiments conducted on benchmark datasets covering three imaging modalities demonstrate that the proposed MSLAU-Net outperforms other state-of-the-art methods on nearly all evaluation metrics, validating the superiority, effectiveness, and robustness of our approach.Our code is available at https://github.com/Monsoon49/MSLAU-Net.

cs.CV↗

arXiv2025

SMRC: Aligning Large Language Models with Student Reasoning for Mathematical Error Correction

Large language models (LLMs) often make reasoning errors when solving mathematical problems, and how to automatically detect and correct these errors has become an important research direction. However, existing approaches \textit{mainly focus on self-correction within the model}, which falls short of the "teacher-style" correction required in educational settings, \textit{i.e.}, systematically guiding and revising a student' s problem-solving process. To address this gap, we propose \texttt{SMRC} (\textit{\underline{S}tudent \underline{M}athematical \underline{R}easoning \underline{C}orrection}), a novel method that aligns LLMs with student reasoning. Specifically, \texttt{SMRC} formulates student reasoning as a multi-step sequential decision problem and introduces Monte Carlo Tree Search (MCTS) to explore optimal correction paths. To reduce the cost of the annotating process-level rewards, we leverage breadth-first search (BFS) guided by LLMs and final-answer evaluation to generate reward signals, which are then distributed across intermediate reasoning steps via a back-propagation mechanism, enabling fine-grained process supervision. Additionally, we construct a benchmark for high school mathematics, MSEB (Multi-Solution Error Benchmark), consisting of 158 instances that include problem statements, student solutions, and correct reasoning steps. We further propose a dual evaluation protocol centered on \textbf{solution accuracy} and \textbf{correct-step retention}, offering a comprehensive measure of educational applicability. Experiments demonstrate that \texttt{SMRC} significantly outperforms existing methods on two public datasets (ProcessBench and MR-GSM8K) and our MSEB in terms of effectiveness and overall performance. The code are available at https://github.com/ECNU-RAIL/SMRC-EMNLP2026.

cs.CL↗

arXiv2025

Large-scale Multi-sequence Pretraining for Generalizable MRI Analysis in Versatile Clinical Applications

Multi-sequence Magnetic Resonance Imaging (MRI) offers remarkable versatility, enabling the distinct visualization of different tissue types. Nevertheless, the inherent heterogeneity among MRI sequences poses significant challenges to the generalization capability of deep learning models. These challenges undermine model performance when faced with varying acquisition parameters, thereby severely restricting their clinical utility. In this study, we present PRISM, a foundation model PRe-trained with large-scale multI-Sequence MRI. We collected a total of 64 datasets from both public and private sources, encompassing a wide range of whole-body anatomical structures, with scans spanning diverse MRI sequences. Among them, 336,476 volumetric MRI scans from 34 datasets (8 public and 26 private) were curated to construct the largest multi-organ multi-sequence MRI pretraining corpus to date. We propose a novel pretraining paradigm that disentangles anatomically invariant features from sequence-specific variations in MRI, while preserving high-level semantic representations. We established a benchmark comprising 44 downstream tasks, including disease diagnosis, image segmentation, registration, progression prediction, and report generation. These tasks were evaluated on 32 public datasets and 5 private cohorts. PRISM consistently outperformed both non-pretrained models and existing foundation models, achieving first-rank results in 39 out of 44 downstream benchmarks with statistical significance improvements. These results underscore its ability to learn robust and generalizable representations across unseen data acquired under diverse MRI protocols. PRISM provides a scalable framework for multi-sequence MRI analysis, thereby enhancing the translational potential of AI in radiology. It delivers consistent performance across diverse imaging protocols, reinforcing its clinical applicability.

eess.IV↗

arXiv2025

Cross-Modal Clustering-Guided Negative Sampling for Self-Supervised Joint Learning from Medical Images and Reports

Learning medical visual representations directly from paired images and reports through multimodal self-supervised learning has emerged as a novel and efficient approach to digital diagnosis in recent years. However, existing models suffer from several severe limitations. 1) neglecting the selection of negative samples, resulting in the scarcity of hard negatives and the inclusion of false negatives; 2) focusing on global feature extraction, but overlooking the fine-grained local details that are crucial for medical image recognition tasks; and 3) contrastive learning primarily targets high-level features but ignoring low-level details which are essential for accurate medical analysis. Motivated by these critical issues, this paper presents a Cross-Modal Cluster-Guided Negative Sampling (CM-CGNS) method with two-fold ideas. First, it extends the k-means clustering used for local text features in the single-modal domain to the multimodal domain through cross-modal attention. This improvement increases the number of negative samples and boosts the model representation capability. Second, it introduces a Cross-Modal Masked Image Reconstruction (CM-MIR) module that leverages local text-to-image features obtained via cross-modal attention to reconstruct masked local image regions. This module significantly strengthens the model's cross-modal information interaction capabilities and retains low-level image features essential for downstream tasks. By well handling the aforementioned limitations, the proposed CM-CGNS can learn effective and robust medical visual representations suitable for various recognition tasks. Extensive experimental results on classification, detection, and segmentation tasks across five downstream datasets show that our method outperforms state-of-the-art approaches on multiple metrics, verifying its superior performance.

cs.CV↗

arXiv2025

Stochastic momentum ADMM for nonconvex and nonsmooth optimization with application to PnP algorithm

This paper proposes SMADMM, a single-loop Stochastic Momentum Alternating Direction Method of Multipliers for solving a class of nonconvex and nonsmooth composite optimization problems. SMADMM achieves the optimal oracle complexity of $\mathcal{O}(ε^{-3/2})$ in the online setting. Unlike previous stochastic ADMM algorithms that require large mini-batches or a double-loop structure, SMADMM uses only $\mathcal{O}(1)$ stochastic gradient evaluations per iteration and avoids costly restarts. To further improve practicality, we incorporate dynamic step sizes and penalty parameters, proving that SMADMM maintains its optimal complexity without the need for large initial batches. We also develop PnP-SMADMM by integrating plug-and-play priors, and establish its theoretical convergence under mild assumptions. Extensive experiments on classification, CT image reconstruction, and phase retrieval tasks demonstrate that our approach outperforms existing stochastic ADMM methods both in accuracy and efficiency, validating our theoretical results.

math.OC↗

arXiv2025

A Large Model for Non-invasive and Personalized Management of Breast Cancer from Multiparametric MRI

Breast Magnetic Resonance Imaging (MRI) demonstrates the highest sensitivity for breast cancer detection among imaging modalities and is standard practice for high-risk women. Interpreting the multi-sequence MRI is time-consuming and prone to subjective variation. We develop a large mixture-of-modality-experts model (MOME) that integrates multiparametric MRI information within a unified structure, leveraging breast MRI scans from 5,205 female patients in China for model development and validation. MOME matches four senior radiologists' performance in identifying breast cancer and outperforms a junior radiologist. The model is able to reduce unnecessary biopsies in Breast Imaging-Reporting and Data System (BI-RADS) 4 patients, classify triple-negative breast cancer, and predict pathological complete response to neoadjuvant chemotherapy. MOME further supports inference with missing modalities and provides decision explanations by highlighting lesions and measuring modality contributions. To summarize, MOME exemplifies an accurate and robust multimodal model for noninvasive, personalized management of breast cancer patients via multiparametric MRI. Code is available at https://github.com/LLYXC/MOME/tree/main.

cs.CV↗

arXiv2024

Inexact Riemannian Gradient Descent Method for Nonconvex Optimization

Gradient descent methods are fundamental first-order optimization algorithms in both Euclidean spaces and Riemannian manifolds. However, the exact gradient is not readily available in many scenarios. This paper proposes a novel inexact Riemannian gradient descent algorithm for nonconvex problems, accompanied by a convergence guarantee. In particular, we establish two inexact gradient conditions on Riemannian manifolds for the first time, enabling precise gradient approximations. Our method demonstrates strong convergence results for both gradient sequences and function values. The global convergence with constructive convergence rates for the sequence of iterates is ensured under the Riemannian Kurdyka-Łojasiewicz property. Furthermore, our algorithm encompasses two specific applications: Riemannian sharpness-aware minimization and Riemannian extragradient algorithm, both of which inherit the global convergence properties of the inexact gradient methods. Numerical experiments on low-rank matrix completion and principal component analysis problems validate the efficiency and practical relevance of the proposed approaches.

math.OC↗

arXiv2024

Image-Based Virtual Try-On: A Survey

Image-based virtual try-on aims to synthesize a naturally dressed person image with a clothing image, which revolutionizes online shopping and inspires related topics within image generation, showing both research significance and commercial potential. However, there is a gap between current research progress and commercial applications and an absence of comprehensive overview of this field to accelerate the development.In this survey, we provide a comprehensive analysis of the state-of-the-art techniques and methodologies in aspects of pipeline architecture, person representation and key modules such as try-on indication, clothing warping and try-on stage. We additionally apply CLIP to assess the semantic alignment of try-on results, and evaluate representative methods with uniformly implemented evaluation metrics on the same dataset.In addition to quantitative and qualitative evaluation of current open-source methods, unresolved issues are highlighted and future research directions are prospected to identify key trends and inspire further exploration. The uniformly implemented evaluation metrics, dataset and collected methods will be made public available at https://github.com/little-misfit/Survey-Of-Virtual-Try-On.

cs.CV↗

arXiv2023

Chiral and nonreciprocal single-photon scattering in a chiral-giant-molecule waveguide-QED system

We study chiral and nonreciprocal single-photon scattering in a chiral-giant-molecule waveguide-QED system. Here, the giant molecule consists of two coupled giant atoms, which interact with two linear waveguides, forming a four-port quantum device. We obtain the exact analytical expressions of the four scattering amplitudes using a real-space method. Under the Markovian limit, we find that the single-photon scattering behavior is determined by the coupling strength between the giant atoms and the waveguides, the coupling strength between the two giant atoms, and the nondipole effect caused by the phase accumulation of photons travelling between the coupling points. It is also found that chiral and nonreciprocal single-photon scattering can be realized by introducing the chiral coupling to break the symmetry in the coupling configuration between the giant molecule and the waveguides. In addition, an ideal chiral emitter-waveguide coupling enables a directional single-photon routing. In the non-Markovian regime, the scattering spectra are characterized by more abundant structures with multiple peaks and dips. In particular, we demonstrate that the non-Markovian retarded effect can induce the nonreciprocal single-photon scattering. Our results have potential applications in the design of optical quantum devices involving giant atoms, which can provide an efficient platform for studying chiral quantum optics.

quant-ph↗

arXiv2020

Achromatic metasurfaces with inversely customized dispersion for ultra-broadband acoustic beam engineering

Metasurfaces, the ultrathin media with extraordinary wavefront modulation ability, have shown versatile potential in manipulating waves. However, existing acoustic metasurfaces are limited by their narrow-band frequency-dependent capability, which severely hinders their real-world applications that usually require customized dispersion. To address this bottlenecking challenge, we report ultra-broadband achromatic metasurfaces that are capable of delivering arbitrary and frequency-independent wave properties by bottom-up topology optimization. We successively demonstrate three ultra-broadband functionalities, including acoustic beam steering, focusing and levitation, featuring record-breaking relative bandwidths of 93.3%, 120% and 118.9%, respectively. All metasurface elements show novel asymmetric geometries containing multiple scatters, curved air channels and local cavities. Moreover, we reveal that the inversely designed metasurfaces can support integrated internal resonances, bi-anisotropy and multiple scattering, which collectively form the mechanism underpinning the ultra-broadband customized dispersion. Our study opens new horizons for ultra-broadband high-efficiency achromatic functional devices on demand, with promising extension to the optical and elastic achromatic metamaterials.

cond-mat.mtrl-sci↗

arXiv2019

Deep Angular Embedding and Feature Correlation Attention for Breast MRI Cancer Analysis

Accurate and automatic analysis of breast MRI plays an important role in early diagnosis and successful treatment planning for breast cancer. Due to the heterogeneity nature, accurate diagnosis of tumors remains a challenging task. In this paper, we propose to identify breast tumor in MRI by Cosine Margin Sigmoid Loss (CMSL) with deep learning (DL) and localize possible cancer lesion by COrrelation Attention Map (COAM) based on the learned features. The CMSL embeds tumor features onto a hypersphere and imposes a decision margin through cosine constraints. In this way, the DL model could learn more separable inter-class features and more compact intra-class features in the angular space. Furthermore, we utilize the correlations among feature vectors to generate attention maps that could accurately localize cancer candidates with only image-level label. We build the largest breast cancer dataset involving 10,290 DCE-MRI scan volumes for developing and evaluating the proposed methods. The model driven by CMSL achieved classification accuracy of 0.855 and AUC of 0.902 on the testing set, with sensitivity and specificity of 0.857 and 0.852, respectively, outperforming other competitive methods overall. In addition, the proposed COAM accomplished more accurate localization of the cancer center compared with other state-of-the-art weakly supervised localization method.

eess.IV↗

arXiv2011

Iron-chalcogenide FeSe$_{0.5}$Te$_{0.5}$ coated superconducting tapes for high field applications

The high upper critical field characteristic of the recently discovered iron-based superconducting chalcogenides opens the possibility of developing a new type of non-oxide high-field superconducting wires. In this work, we utilize a buffered metal template on which we grow a textured FeSe$_{0.5}$Te$_{0.5}$ layer, an approach developed originally for high temperature superconducting coated conductors. These tapes carry high critical current densities (>1$\times10^{4}$A/cm$^{2}$) at about 4.2K under magnetic field as high as 25 T, which are nearly isotropic to the field direction. This demonstrates a very promising future for iron chalcogenides for high field applications at liquid helium temperatures. Flux pinning force analysis indicates a point defect pinning mechanism, creating prospects for a straightforward approach to conductor optimization.

cond-mat.supr-con↗

arXiv2010

Nanostructures and Defects in Non-equilibrium Synthesized Filled Skutterudite CeFe4Sb12

We studied nanoprecipitates and defects in p-type filled skutterudite CeFe4Sb12 prepared by non-equilibrium melt-spinning plus spark plasma sintering method using transmission electron microscopy. Nanoprecipitates with mostly spherical shapes and different sizes (from several nm to several tens of nm) have been observed. The most typically observed nanoprecipitates are shown to be Sb-rich. Superlattices with a periodicity of about 3.576 nm were induced by the ordering of excessive Sb atoms along the c direction. These nanoprecipitates usually share coherent interfaces with the surrounding matrix and induce anisotropic and strong strain fields in the surrounding matrix. Nanoprecipitates with compositions close to CeSb2 are much larger in size (~ 30 nm) and have orthorhombic structures. Various defects were typically observed on the interfaces between these nanoprecipitates and the matrix. The strain fields induced by these nanoprecipitates are less distinct, possibly because part of the strains has been released by the formation of defects.

cond-mat.mtrl-sci↗

arXiv2010

Strong impact of grain boundaries on the thermoelectric properties of non-equilibrium synthesized p-type Ce1.05Fe4Sb12.04 filled skutterudites with nanostructure

p-type Ce1.05Fe4Sb12.04 filled skutterudites with much improved thermoelectric properties have been synthesized by rapidly converting nearly amorphous ribbons into crystalline pellets under pressure. It is found that this process greatly suppresses grain growth and second phase formation/segregation, and hence results in the samples consisting of nano-sized grains with strongly-coupled grain boundaries, as observed by transmission electron microscopy. The room temperature carrier mobility in these samples is significantly higher (nearly double) than those in the samples of the same starting composition made by the conventional solid-state reaction. Nanostructure reduces the lattice thermal conductivity, while cleaner grain boundaries permit higher electron conduction.

cond-mat.mtrl-sci↗