arXiv subjects

Xinyu Chen

Publications and source records attributed to Xinyu Chen.

At least 55 records · Page 3Linked to original sources

arXiv2026

INSPATIO-WORLD: A Real-Time 4D World Simulator via Spatiotemporal Autoregressive Modeling

Building world models with spatial consistency and real-time interactivity remains a fundamental challenge in computer vision. Current video generation paradigms often struggle with a lack of spatial persistence and insufficient visual realism, making it difficult to support seamless navigation in complex environments. To address these challenges, we propose INSPATIO-WORLD, a novel real-time framework capable of recovering and generating high-fidelity, dynamic interactive scenes from a single reference video. At the core of our approach is a Spatiotemporal Autoregressive (STAR) architecture, which enables consistent and controllable scene evolution through two tightly coupled components: Implicit Spatiotemporal Cache aggregates reference and historical observations into a latent world representation, ensuring global consistency during long-horizon navigation; Explicit Spatial Constraint Module enforces geometric structure and translates user interactions into precise and physically plausible camera trajectories. Furthermore, we introduce Joint Distribution Matching Distillation (JDMD). By using real-world data distributions as a regularizing guide, JDMD effectively overcomes the fidelity degradation typically caused by over-reliance on synthetic data. Extensive experiments demonstrate that INSPATIO-WORLD significantly outperforms existing state-of-the-art (SOTA) models in spatial consistency and interaction precision, ranking first among real-time interactive methods on the WorldScore-Dynamic benchmark, and establishing a practical pipeline for navigating 4D environments reconstructed from monocular videos.

cs.CV↗

arXiv2026

FashionStylist: An Expert Knowledge-enhanced Multimodal Dataset for Fashion Understanding

Fashion understanding requires both visual perception and expert-level reasoning about style, occasion, compatibility, and outfit rationale. However, existing fashion datasets remain fragmented and task-specific, often focusing on item attributes, outfit co-occurrence, or weak textual supervision, and thus provide limited support for holistic outfit understanding. In this paper, we introduce FashionStylist, an expert-annotated benchmark for holistic and expert-level fashion understanding. Constructed through a dedicated fashion-expert annotation pipeline, FashionStylist provides professionally grounded annotations at both the item and outfit levels. It supports three representative tasks: outfit-to-item grounding, outfit completion, and outfit evaluation. These tasks cover realistic item recovery from complex outfits with layering and accessories, compatibility-aware composition beyond co-occurrence matching, and expert-level assessment of style, season, occasion, and overall coherence. Experimental results show that FashionStylist serves not only as a unified benchmark for multiple fashion tasks, but also as an effective training resource for improving grounding, completion, and outfit-level semantic evaluation in MLLM-based fashion systems.

cs.CV↗

arXiv2026

InfiniLoRA: Disaggregated Multi-LoRA Serving for Large Language Models

LoRA enables efficient customization of LLMs and is widely used in multi-tenant and multi-task serving. However, emerging model architectures such as MoE significantly increase LoRA memory cost, making existing coupled LoRA serving designs poorly scalable and prone to tail-latency inflation. We present InfiniLoRA, a disaggregated LoRA serving system that decouples LoRA execution from base-model inference. InfiniLoRA introduces a shared LoRA Server with parallelism-aware execution, SLO-driven provisioning, and critical-path optimizations, including GPU-initiated communication and hardware-specialized LoRA kernels. Experiments show that InfiniLoRA can achieve an average $3.05\times$ increase in serviceable request rate under strict latency SLOs, and improve the percentage of LoRA adapters satisfying the SLO requirement by 54.0\%.

cs.DC↗

arXiv2026

$α$-robust utility maximization with intractable claims: A quantile optimization approach

This paper studies an $α$-robust utility maximization problem where an investor faces an intractable claim -- an exogenous contingent claim with known marginal distribution but unspecified dependence structure with financial market returns. The $α$-robust criterion interpolates between worst-case ($α=0$) and best-case ($α=1$) evaluations, generalizing both extremes through a continuous ambiguity attitude parameter. For weighted exponential utilities, we establish via rearrangement inequalities and comonotonicity theory that the $α$-robust risk measure is law-invariant, depending only on marginal distributions. This transforms the dynamic stochastic control problem into a concave static quantile optimization over a convex domain. We derive optimality conditions via calculus of variations and characterize the optimal quantile as the solution to a two-dimensional first-order ordinary differential equation system, which is a system of variational inequalities with mixed boundary conditions, enabling numerical solution. Our framework naturally accommodates additional risk constraints such as Value-at-Risk and Expected Shortfall. Numerical experiments reveal how ambiguity attitude, market conditions, and claim characteristics interact to shape optimal payoffs.

q-fin.PM↗

arXiv2026

MSVBench: Towards Human-Level Evaluation of Multi-Shot Video Generation

The evolution of video generation toward complex, multi-shot narratives has exposed a critical deficit in current evaluation methods. Existing benchmarks remain anchored to single-shot paradigms, lacking the comprehensive story assets and cross-shot metrics required to assess long-form coherence and appeal. To bridge this gap, we introduce MSVBench, the first comprehensive benchmark featuring hierarchical scripts and reference images tailored for Multi-Shot Video generation. We propose a hybrid evaluation framework that synergizes the high-level semantic reasoning of Large Multimodal Models (LMMs) with the fine-grained perceptual rigor of domain-specific expert models. Evaluating 20 video generation methods across diverse paradigms, we find that current models--despite strong visual fidelity--primarily behave as visual interpolators rather than true world models. We further validate the reliability of our benchmark by demonstrating a state-of-the-art Spearman's rank correlation of 94.4% with human judgments. Finally, MSVBench extends beyond evaluation by providing a scalable supervisory signal. Fine-tuning a lightweight model on its pipeline-refined reasoning traces yields human-aligned performance comparable to commercial models like Gemini-2.5-Flash.

cs.MM↗

arXiv2026

Sinh regularized Lagrangian nonuniform sampling series

Recently, some window functions have been introduced into the nonuniform fast Fourier transform and the regularized Shannon sampling. Inspired by these works, we utilize a sinh-type function to accelerate the convergence of the Lagrangian nonuniform sampling series. Our theoretical error estimates and numerical experiments demonstrate that the sinh regularized nonuniform sampling series achieves a superior convergence rate compared to the fastest existing Gaussian regularized nonuniform sampling series.

math.NA↗

arXiv2026

Unveiling the impact of anti-site defects in magnetic transitions of few-layer MnBi2Te4 by operando heating

As the first experimentally discovered intrinsic magnetic topological insulator, MnBi2Te4 has attracted widespread attentions, providing a unique platform for the exploration of topological quantum phases, such as quantum anomalous Hall effect and axion insulator state. Despite the increasing number of potential factors affecting samples being identified, obtaining the high-quality device performance with desired topological quantum phases remains a challenge. In this work, by comparing the reflective magnetic circular dichroism (RMCD) of crystals with different defect densities that are characterized by atomically resolved scanning tunneling microscopy, we demonstrate that anti-site defects play an essential role in achieving ideal magnetic states. By measuring RMCD hysteresis loops with operando heating, we find that MnBi2Te4 few-layer samples are highly susceptible to thermal impact, even at temperature as low as 45°C. The magnetic behavior of heating-treated samples is akin to that of samples fabricated into devices, revealing the thermal impact on devices as well. Starting from few-layers with ideal layer-dependent magnetic order, thermal heating leads to the convergence of magnetization and transition fields between odd- and even-layers. The observed heating-induced magnetic evolution can serve as a valuable reference for assessing the sample quality or the density of anti-site defects. Our findings not only point out the long-standing hidden factor that arose controversies in MnBi2Te4, but also pave the way for controllably engineering the topological quantum phenomena.

cond-mat.mtrl-sci↗

arXiv2026

RidgeWalker: Perfectly Pipelined Graph Random Walks on FPGAs

Graph Random Walks (GRWs) offer efficient approximations of key graph properties and have been widely adopted in many applications. However, GRW workloads are notoriously difficult to accelerate due to their strong data dependencies, irregular memory access patterns, and imbalanced execution behavior. While recent work explores FPGA-based accelerators for GRWs, existing solutions fall far short of hardware potential due to inefficient pipelining and static scheduling. This paper presents RidgeWalker, a high-performance GRW accelerator designed for datacenter FPGAs. The key insight behind RidgeWalker is that the Markov property of GRWs allows decomposition into stateless, fine-grained tasks that can be executed out-of-order without compromising correctness. Building on this, RidgeWalker introduces an asynchronous pipeline architecture with a feedback-driven scheduler grounded in queuing theory, enabling perfect pipelining and adaptive load balancing. We prototype RidgeWalker on datacenter FPGAs and evaluated it across a range of GRW algorithms and real-world graph datasets. Experimental results demonstrate that RidgeWalker achieves an average speedup of 7.0x over state-of-the-art FPGA solutions and 8.1x over GPU solutions, with peak speedups of up to 71.0x and 22.9x, respectively. The source code is publicly available at https://github.com/Xtra-Computing/RidgeWalker.

cs.AR↗

arXiv2025

Symplectic methods for stochastic Hamiltonian systems: asymptotic error distributions and Hamiltonian-specific analysis

In this paper, we investigate the asymptotic error distributions of symplectic methods for stochastic Hamiltonian systems and further provide Hamiltonian-specific analysis that clarifies the superiority of symplectic methods. Our contribution is threefold. First, we derive the asymptotic error distributions of symplectic methods for stochastic Hamiltonian systems with multiplicative noise and additive noise, respectively, and show that the obtained limiting stochastic processes satisfy equations retaining the Hamiltonian formulations. Second, we propose a new approach for calculating the asymptotic error distribution, revealing the connection between the stochastic modified equation and the asymptotic error distribution. Third, we characterize the limiting distribution of the normalized Hamiltonian deviation, thereby illustrating through test equations the superiority of symplectic methods for long-time simulations of the Hamiltonians, even in the limit as the step size tends to zero.

math.NA↗

arXiv2025

Uni-MoE-2.0-Omni: Scaling Language-Centric Omnimodal Large Model with Advanced MoE, Training and Data

We present Uni-MoE 2.0 from the Lychee family. As a fully open-source omnimodal large model (OLM), it substantially advances Lychee's Uni-MoE series in language-centric multimodal understanding, reasoning, and generating. Based on the dense LLM, we build Uni-MoE-2.0-Omni from scratch through three core contributions: dynamic-capacity Mixture-of-Experts (MoE) design, a progressive training strategy enhanced with an iterative reinforcement strategy, and a carefully curated multimodal data matching technique. It is capable of omnimodal understanding, as well as generating images, text, and speech. Architecturally, our new MoE framework balances computational efficiency and capability for 10 cross-modal inputs using shared, routed, and null experts, while our Omni-Modality 3D RoPE ensures spatio-temporal cross-modality alignment in the self-attention layer. For training, following cross-modal pretraining, we use a progressive supervised fine-tuning strategy that activates modality-specific experts and is enhanced by balanced data composition and an iterative GSPO-DPO method to stabilise RL training and improve reasoning. Data-wise, the base model, trained on approximately 75B tokens of open-source multimodal data, is equipped with special speech and image generation tokens, allowing it to learn these generative tasks by conditioning its outputs on linguistic cues. Extensive evaluation across 85 benchmarks demonstrates that our model achieves SOTA or highly competitive performance against leading OLMs, surpassing Qwen2.5-Omni (trained with 1.2T tokens) on over 50 of 76 benchmarks. Key strengths include video understanding (+7% avg. of 8), omnimodallity understanding (+7% avg. of 4), and audiovisual reasoning (+4%). It also advances long-form speech processing (reducing WER by 4.2%) and leads in low-level image processing and controllable generation across 5 metrics.

cs.CL↗

arXiv2025

Provably Minimum-Length Conformal Prediction Sets for Ordinal Classification

Ordinal classification has been widely applied in many high-stakes applications, e.g., medical imaging and diagnosis, where reliable uncertainty quantification (UQ) is essential for decision making. Conformal prediction (CP) is a general UQ framework that provides statistically valid guarantees, which is especially useful in practice. However, prior ordinal CP methods mainly focus on heuristic algorithms or restrictively require the underlying model to predict a unimodal distribution over ordinal labels. Consequently, they provide limited insight into coverage-efficiency trade-offs, or a model-agnostic and distribution-free nature favored by CP methods. To this end, we fill this gap by propose an ordinal-CP method that is model-agnostic and provides instance-level optimal prediction intervals. Specifically, we formulate conformal ordinal classification as a minimum-length covering problem at the instance level. To solve this problem, we develop a sliding-window algorithm that is optimal on each calibration data, with only a linear time complexity in K, the number of label candidates. The local optimality per instance further also improves predictive efficiency in expectation. Moreover, we propose a length-regularized variant that shrinks prediction set size while preserving coverage. Experiments on four benchmark datasets from diverse domains are conducted to demonstrate the significantly improved predictive efficiency of the proposed methods over baselines (by 15% decrease on average over four datasets).

cs.LG↗

arXiv2025

Broadband nonlinear Hall response and multiple wave mixing in a room temperature altermagnet

Crystalline symmetries determine the linear and nonlinear response of materials to external stimuli such as mechanical pressure and electromagnetic fields, governing phenomena such as piezoelectricity, optical activity, and multiple wave mixing with wide ranging technological applications. Altermagnets present a new class of materials with magnetic crystalline order where specific crystal symmetry operations connect antiferromagnetic sublattices, leading to non-relativistic spin-splitting of the electronic band structure. Hence, the electric material properties of altermagnets should uniquely mirror these underlying symmetry properties, potentially giving rise to novel phenomena in response to external driving fields. Here, we report the discovery of a broadband third-order nonlinear anomalous Hall effect in altermagnetic CrSb at room temperature. The comparison of our observations with symmetry analyses and model calculations shows that this nonlinear Hall response is induced by the nonlinear electric susceptibility of a Berry curvature quadrupole, which exists within the spin-split band structure of CrSb and is characterized by the underlying crystalline and magnetic symmetries. We then utilize this third-order nonlinear electric susceptibility of CrSb to realize a multiple wave mixing device with pronounced four wave mixing output, which could, in principle, be extended to THz frequencies. Our study discovers that the crystalline magnetic order of altermagnets determines their nonlinear electric material properties, which could facilitate applications in high-frequency electronics, THz generation, communication networks, and energy harvesting.

cond-mat.mes-hall↗

arXiv2025

Knowledge-Guided Prompt Learning for Request Quality Assurance in Public Code Review

Public Code Review (PCR) is developed in the Software Question Answering (SQA) community, assisting developers in exploring high-quality and efficient review services. Current methods on PCR mainly focus on the reviewer's perspective, including finding a capable reviewer, predicting comment quality, and recommending/generating review comments. However, it is not well studied that how to satisfy the review necessity requests posted by developers which can increase their visibility, which in turn acts as a prerequisite for better review responses. To this end, we propose K nowledge-guided P rompt learning for P ublic Code Review (KP-PCR) to achieve developer-based code review request quality assurance (i.e., predicting request necessity and recommending tags subtask). Specifically, we reformulate the two subtasks via 1) text prompt tuning which converts both of them into a Masked Language Model (MLM) by constructing prompt templates using hard prompt; and 2) knowledge and code prefix tuning which introduces knowledge guidance from fine-tuned large language models by soft prompt, and uses program dependence graph to characterize code snippets. Finally, both of the request necessity prediction and tag recommendation subtasks output predicted results through an answer engineering module. In addition, we further analysis the time complexity of our KP-PCR that has lightweight prefix based the operation of introducing knowledge guidance. Experimental results on the PCR dataset for the period 2011-2023 demonstrate that our KP-PCR outperforms baselines by 2.3%-8.4% in the request necessity prediction and by 1.4%-6.9% in the tag recommendation. The code implementation is released at https://github.com/WUT-IDEA/KP-PCR

cs.SE↗

arXiv2025

Medical Referring Image Segmentation via Next-Token Mask Prediction

Medical Referring Image Segmentation (MRIS) involves segmenting target regions in medical images based on natural language descriptions. While achieving promising results, recent approaches usually involve complex design of multimodal fusion or multi-stage decoders. In this work, we propose NTP-MRISeg, a novel framework that reformulates MRIS as an autoregressive next-token prediction task over a unified multimodal sequence of tokenized image, text, and mask representations. This formulation streamlines model design by eliminating the need for modality-specific fusion and external segmentation models, supports a unified architecture for end-to-end training. It also enables the use of pretrained tokenizers from emerging large-scale multimodal models, enhancing generalization and adaptability. More importantly, to address challenges under this formulation-such as exposure bias, long-tail token distributions, and fine-grained lesion edges-we propose three novel strategies: (1) a Next-k Token Prediction (NkTP) scheme to reduce cumulative prediction errors, (2) Token-level Contrastive Learning (TCL) to enhance boundary sensitivity and mitigate long-tail distribution effects, and (3) a memory-based Hard Error Token (HET) optimization strategy that emphasizes difficult tokens during training. Extensive experiments on the QaTa-COV19 and MosMedData+ datasets demonstrate that NTP-MRISeg achieves new state-of-the-art performance, offering a streamlined and effective alternative to traditional MRIS pipelines.

cs.CV↗

arXiv2025

The law of iterated logarithm for numerical approximation of time-homogeneous Markov process

The law of the iterated logarithm (LIL) for the time-homogeneous Markov process with a unique invariant measure characterizes the almost sure maximum possible fluctuation of time averages around the ergodic limit. Whether a numerical approximation can preserve this asymptotic pathwise behavior remains an open problem. In this work, we give a positive answer to this question and establish the LIL for the numerical approximation of such a process under verifiable assumptions. The Markov process is discretized by a decreasing time-step strategy, which yields the non-homogeneous numerical approximation but facilitates a martingale-based analysis. The key ingredient in proving the LIL for such numerical approximation lies in extracting a quasi-uniform time-grid subsequence from the original non-uniform time grids and establishing the LIL for a predominant martingale along it, while the remainder terms converge to zero. Finally, we illustrate that our results can be flexibly applied to numerical approximations of a broad class of stochastic systems, including SODEs and SPDEs.

math.NA↗

arXiv2025

FreeControl: Efficient, Training-Free Structural Control via One-Step Attention Extraction

Controlling the spatial and semantic structure of diffusion-generated images remains a challenge. Existing methods like ControlNet rely on handcrafted condition maps and retraining, limiting flexibility and generalization. Inversion-based approaches offer stronger alignment but incur high inference cost due to dual-path denoising. We present FreeControl, a training-free framework for semantic structural control in diffusion models. Unlike prior methods that extract attention across multiple timesteps, FreeControl performs one-step attention extraction from a single, optimally chosen key timestep and reuses it throughout denoising. This enables efficient structural guidance without inversion or retraining. To further improve quality and stability, we introduce Latent-Condition Decoupling (LCD): a principled separation of the key timestep and the noised latent used in attention extraction. LCD provides finer control over attention quality and eliminates structural artifacts. FreeControl also supports compositional control via reference images assembled from multiple sources - enabling intuitive scene layout design and stronger prompt alignment. FreeControl introduces a new paradigm for test-time control, enabling structurally and semantically aligned, visually coherent generation directly from raw images, with the flexibility for intuitive compositional design and compatibility with modern diffusion models at approximately 5 percent additional cost.

cs.CV↗

arXiv2025

Transfer learning discovery of molecular modulators for perovskite solar cells

The discovery of effective molecular modulators is essential for advancing perovskite solar cells (PSCs), but the research process is hindered by the vastness of chemical space and the time-consuming and expensive trial-and-error experimental screening. Concurrently, machine learning (ML) offers significant potential for accelerating materials discovery. However, applying ML to PSCs remains a major challenge due to data scarcity and limitations of traditional quantitative structure-property relationship (QSPR) models. Here, we apply a chemical informed transfer learning framework based on pre-trained deep neural networks, which achieves high accuracy in predicting the molecular modulator's effect on the power conversion efficiency (PCE) of PSCs. This framework is established through systematical benchmarking of diverse molecular representations, enabling lowcost and high-throughput virtual screening over 79,043 commercially available molecules. Furthermore, we leverage interpretability techniques to visualize the learned chemical representation and experimentally characterize the resulting modulator-perovskite interactions. The top molecular modulators identified by the framework are subsequently validated experimentally, delivering a remarkably improved champion PCE of 26.91% in PSCs.

cond-mat.mtrl-sci↗

arXiv2025

AnTKV: Anchor Token-Aware Sub-Bit Vector Quantization for KV Cache in Large Language Models

Quantization has emerged as an effective and lightweight solution to reduce the memory footprint of the KV cache in Large Language Models. Nevertheless, minimizing the accuracy degradation caused by ultra-low-bit KV cache quantization remains a significant challenge. While scalar quantization is constrained by 1-bit bound, vector quantization exploits intra-vector correlations and enables sub-bit regimes, making it more suitable for ultra-low-bit quantization. To further mitigate quantization-induced degradation, we reveal that the degradation is highly uneven across tokens in attention quality. To investigate this unevenness, we introduce anchor score to measure each token's sensitivity to quantization. Our analysis and experiments show that preserving a small subset (1\%) of tokens with the highest Anchor Score significantly mitigates accuracy loss under aggressive quantization. We propose AnTKV, a dual-stage framework that leverages anchor token-aware vector quantization to compress the KV cache. It combines offline token-aware centroids learning and online anchor token selection to balance compression and accuracy. To enable efficient deployment, we design an online anchor token selection kernel compatible with FlashAttention. It allows LLaMA3-8B to scale to 840K tokens on a single 80GB A100, while delivering up to $3.5\times$ higher decoding throughput over the FP16 baseline. Experiments demonstrate that AnTKV matches or surpasses prior methods at 4-bit, and significantly reduce perplexity under ultra-low-bit quantization, achieving 6.32 at 1-bit on Mistral-7B, compared to 7.25 for CQ and 15.36 for KVQuant.

cs.CL↗