arXiv subjects

Yutao Ma

Publications and source records attributed to Yutao Ma.

At least 19 recordsLinked to original sources

arXiv2026

When Fusion Fails: Corruption-Aware Rebalanced Fusion for Multi-Modal Medical Image Segmentation

Multi-modal medical image segmentation leverages complementary diagnostic information, yet fusion can underperform single-modality baselines when spatially aligned inputs differ in quality. Here, "corruption" primarily denotes resolution-induced degradation rather than misalignment or complete modality absence, while synthetic noise is evaluated only as an auxiliary setting. We identify a critical optimization-inference inconsistency: degraded modalities can receive weak training updates yet substantially affect predictions, indicating active interference with fusion. We attribute this failure to resampling-induced feature corruption and optimization bias, where noisy features propagate through skip connections and encourage unreliable modality selection. We therefore propose CoReFuse-Med, a Corruption-aware Rebalanced Fusion framework that suppresses corruption during feature transmission and rebalances modality contributions during high-level fusion. Experiments on EPVS, BraTS, and WMH, including multiple Z-axis slice-retention ratios and an auxiliary noise test, demonstrate improved accuracy and robustness under modality-quality discrepancies. Our code is available at https://github.com/lrever/CoReFuse.

cs.CV↗

arXiv2026

Precise universal edge asymptotics for planar $\beta=2$ Coulomb gases with radial external fields

We investigate the extremal statistics of planar $\beta=2$ Coulomb gases with radial external fields. For the rightmost eigenvalue and the spectral radius, we establish sharp Berry--Esseen bounds for their convergence to the Gumbel distribution, with explicit rates \[ \frac{25\log\log n}{4e\log n} \quad\text{and}\quad \frac{2\log\log n}{e\log n}, \] respectively. In addition, we derive sharp asymptotic equivalences for the large and moderate deviations of both statistics across all relevant scales. Analogous results hold for the smallest modulus.

math.PR↗

arXiv2026

Fine-Tuning Code Language Models to Detect Cross-Language Bugs

Multilingual programming, which involves using multiple programming languages (PLs) in a single project, is increasingly common due to its benefits. However, it introduces cross-language bugs (CLBs), which arise from interactions between different PLs and are difficult to detect by single-language bug detection tools. This paper investigates the potential of pre-trained code language models (CodeLMs) in CLB detection. We developed CLCFinder, a cross-language code identification tool, and constructed a CLB dataset involving three PL combinations (Python-C/C++, Java-C/C++, and Python-Java) with nine interaction types. We fine-tuned 13 CodeLMs on this dataset and evaluated their performance, analyzing the effects of dataset size, token sequence length, and code comments. Results show that all 13 CodeLMs exhibited varying degrees of performance improvement after fine-tuning, with UniXcoder-base achieving the best F1 score (0.7407). Notably, within our experimental setup, small CodeLMs tended to performe better than large ones. CodeLMs fine-tuned on single-language bug datasets performed poorly on CLB detection, demonstrating the distinction between CLBs and single-language bugs. Additionally, increasing the fine-tuning dataset size significantly improved performance, while longer token sequences did not necessarily improve the model performance. The impact of code comments varied across models. Some fine-tuned CodeLMs' performance was improved, while others showed degraded performance.

cs.SE↗

arXiv2026

From Language to Action: Enhancing LLM Task Efficiency with Task-Aware MCP Server Recommendation

The rapid expansion of the model context protocol (MCP) ecosystem enables large language model (LLM)-based agents to access a wide range of external tools via a standardized interface. However, identifying appropriate MCP servers for a specific development task remains challenging. Existing studies primarily focus on measuring the MCP ecosystem or optimizing tool invocation mechanisms, while systematic recommendation frameworks and reproducible benchmarks for real-world development tasks remain largely unexplored. To address this limitation, we formulate task-oriented MCP server recommendation as a structured retrieval-and-ranking problem that jointly considers semantic relevance and engineering constraints. We first construct Task2MCP, a task-centered dataset that systematically associates taxonomy-grounded development tasks with curated MCP servers. This dataset provides structured supervision and a reproducible evaluation environment for research on MCP tool recommendations. Building on this dataset, we propose T2MRec, a task-to-MCP server recommendation model. It models semantic relevance and structural compatibility to construct an initial candidate set. Then it improves coverage and ranking quality through centroid-based candidate expansion and constrained LLM-based re-ranking. In addition, we design and implement an interactive MCP server recommendation agent prototype that operates in conversational environments to support dynamic decision-making. The agent assists developers in efficiently evaluating and integrating tools by providing recommended MCP servers together with usage guidelines.

cs.SE↗

arXiv2026

From Gaussian to Gumbel: extreme eigenvalues of complex Ginibre products with exact rates

We consider the product of $k_{n}$ independent $n\times n$ complex Ginibre matrices and denote its eigenvalues by $Z_{1},\ldots ,Z_{n}$. Let $α= \lim_{n\to\infty} n / k_{n}$. Using the determinantal point process method, we reduce the study of extremal eigenvalues to the evaluation of determinants of certain $n\times n$ matrices. In the modulus case, rotational invariance makes the relevant matrix diagonal, which yields a product representation in terms of Gamma tail probabilities. In the real-part case, the matrix is no longer diagonal; we handle this by a polar-coordinate reduction that introduces an independent uniform angle and leads to explicit formulas involving Gamma variables and trigonometric integrals. After appropriate rescaling, the spectral radius $\max_{1\leq j\leq n}|Z_{j}|$ converges weakly to a nontrivial distribution $Φ_α$ when $α\in (0, +\infty)$, to the Gumbel distribution when $α= +\infty$, and to the standard normal distribution when $α= 0$. The family $\{Φ_α\}_{α>0}$ extends continuously to the boundary regimes: $Φ_α$ converges weakly to the standard normal law as $α\to 0^{+}$ and to the Gumbel law as $α\to +\infty$. Thus the three limiting regimes are connected by the single parameter $α$, yielding a continuous transition from Gaussian to Gumbel distribution. For the spectral radius, we obtain the exact rates of convergence both in the fixed-$α$ regime and at the boundaries $α= 0$ and $α= +\infty$. For the rightmost eigenvalue $\max_{1\leq j\leq n}\Re Z_{j}$, we establish the convergence rates in the boundary regimes, while for $α\in (0, +\infty)$ we show that the limiting distribution, though not available in closed form, still interpolates continuously between the normal and Gumbel laws.

math.PR↗

arXiv2026

Precise convergence rate of spectral radius of product of complex Ginibre

Let $Z_1, \cdots, Z_n$ denote the eigenvalues of the product $\prod_{j=1}^{k_n} \boldsymbol{A}_j$, where $\{\boldsymbol{A}_j\}_{1 \le j \le k_n}$ are independent $n\times n$ complex Ginibre matrices. Define $α= \lim\limits_{n \to \infty} \frac{n}{k_n}$. We prove that $X_n,$ a suitably rescaled version of $\max_{1 \le j \le n} |Z_j|^2,$ converges weakly as follows: to a non-trivial distribution $Φ_α$ for $α\in (0, +\infty)$, to the Gumbel distribution when $α= +\infty$, and to the standard normal distribution when $α= 0$. This result reveals a phase transition at the boundaries of $α$. Furthermore, we establish the exact rates of convergence in each regime.

math.PR↗

arXiv2026

Unveiling Code Clones in the Eclipse IIoT Software Ecosystem

Industrial Internet of Things (IIoT) has become a prominent topic recently, with an increasing number of IIoT OSS projects emerging, also within the Eclipse Foundation. Code cloning is a common practice that can adversely affect software maintenance. In the IIoT OSS domain, developers frequently reuse code and configurations for efficiency, which can lead to code clone proliferation and maintenance challenges. However, the extent and effects of code clones in the IIoT OSS domain remain understudied. This study aims to investigate the prevalence, evolution, and co-modification of code clones within the Eclipse IIoT OSS ecosystem. We collected 90 release versions from 15 projects in the Eclipse IIoT OSS ecosystem, and investigated their code clone situations based on source code and change history using the NiCad tool and our custom analysis module. The investigation covered clone distribution, patterns, evolution trends, co-modified clones, and cross-project clones. 1) Code clones are prevalent in Eclipse IIoT OSS projects, with 16.3% of code lines involved in clones - nearly twice the proportion observed in traditional OSS projects; 2) Most code clones occur between commits, while there are still a significant proportion of code clones that each clone pair happens within a commit; 3) Most Eclipse IIoT projects remain stable in clone numbers during version iterations; 4) An average of 0.17% of the clones have been co-modified, which negatively affect maintenance; and 5) Cross-project clone pairs are prevalent, more in Java than in C projects, with rare co-modifications (0.02%) only in Java projects. Our findings highlight the potential negative impacts of these clones on software maintenance, emphasizing the need to address these issues to improve overall software quality.

cs.SE↗

arXiv2026

Deviation probabilities and Sharp Berry-Esseen bound for rightmost eigenvalue of large non-Hermitian chiral random matrices

This paper provides a quantitative analysis of the rightmost eigenvalue for a chiral non-Hermitian random Dirac matrix in the maximally non-Hermitian regime ($τ=0$). Let $(σ_i)_{1\le i\le n}$ be the eigenvalues with positive real part. We define the normalization constants \[ s_n = \frac{4n(n+v)}{2n+v}, \qquad γ_n = \frac{1}{2}\log s_n - \frac{5}{4}\log(\log s_n) - \log\bigl(2^{1/4}π\bigr), \] and the centered and scaled variable \[ X_n = \sqrt{2s_n\log s_n}\,\bigl(\bigl(\tfrac{n}{n+v}\bigr)^{1/4}\,\max_{1\le i\le n}\Reσ_i \;-\; 1 \;-\; \frac{γ_n}{\sqrt{2s_n\log s_n}}\bigr). \] Our main result is the following sharp Berry--Esseen bound for the convergence of $X_n$ to the Gumbel distribution: \[ \sup_{x \in \mathbb{R}} \bigl|\mathbb{P}(X_n \le x) - e^{-e^{-x}}\bigr| = \frac{25 (\log\log s_n)^2}{16 e \,\log s_n}\,\bigl(1 + o(1)\bigr), \] which holds as $n \to \infty$ for an arbitrary parameter $v \ge 0$ (which may depend on $n$). As a byproduct of our analysis, we also obtain precise large- and moderate-deviation principles for the scaled rightmost eigenvalue $\bigl(\frac{n}{n+v}\bigr)^{1/4} \max_{1\le i\le n}\Reσ_i$, characterizing its rate of convergence to the value $1$.

math.PR↗

arXiv2025

Standardized Evaluation of Automatic Methods for Perivascular Spaces Segmentation in MRI -- MICCAI 2024 Challenge Results

Perivascular spaces (PVS), when abnormally enlarged and visible in magnetic resonance imaging (MRI) structural sequences, are important imaging markers of cerebral small vessel disease and potential indicators of neurodegenerative conditions. Despite their clinical significance, automatic enlarged PVS (EPVS) segmentation remains challenging due to their small size, variable morphology, similarity with other pathological features, and limited annotated datasets. This paper presents the EPVS Challenge organized at MICCAI 2024, which aims to advance the development of automated algorithms for EPVS segmentation across multi-site data. We provided a diverse dataset comprising 100 training, 50 validation, and 50 testing scans collected from multiple international sites (UK, Singapore, and China) with varying MRI protocols and demographics. All annotations followed the STRIVE protocol to ensure standardized ground truth and covered the full brain parenchyma. Seven teams completed the full challenge, implementing various deep learning approaches primarily based on U-Net architectures with innovations in multi-modal processing, ensemble strategies, and transformer-based components. Performance was evaluated using dice similarity coefficient, absolute volume difference, recall, and precision metrics. The winning method employed MedNeXt architecture with a dual 2D/3D strategy for handling varying slice thicknesses. The top solutions showed relatively good performance on test data from seen datasets, but significant degradation of performance was observed on the previously unseen Shanghai cohort, highlighting cross-site generalization challenges due to domain shift. This challenge establishes an important benchmark for EPVS segmentation methods and underscores the need for the continued development of robust algorithms that can generalize in diverse clinical settings.

q-bio.QM↗

arXiv2025

Bug Priority Change Prediction: An Exploratory Study on Apache Software

Bug fixing is a critical activity in the software development process. In issue tracking systems such as JIRA, each bug report is assigned a priority level to indicate the urgency and importance level of the bug. The priority may change during the bug fixing process, indicating that the urgency and importance level of the bug will change with the bug fixing. However, manually evaluating priority changes for bugs is a tedious process that heavily relies on the subjective judgment of developers and project managers, leading to incorrect priority changes and thus hindering timely bug fixes. Given the lack of research on bug priority change prediction, we propose a novel two-phase bug report priority change prediction method based on bug fixing evolution features and class imbalance handling strategy. Specifically, we divided the bug lifecycle into two phases: bug reporting and bug fixing, and constructed bug priority change prediction models for each phase. To evaluate the performance of our method, we conducted experiments on a bug dataset constructed from 32 non-trivial Apache projects. The experimental results show that our proposed bug fixing evolution features and the adopted class imbalance handling strategy can effectively improve the performance of prediction models. The F1-score of the prediction model constructed for the bug reporting phase reached 0.798, while the F1-weighted and F1-macro of the prediction model constructed for the bug fixing phase were 0.712 and 0.613, respectively. Furthermore, we explored the cross-project applicability of our prediction models and their performance at different priority levels. The findings indicate large variations in model performance across different projects, although the overall scores remain decent. Meanwhile, the predictive performance across various priority levels remained relatively consistently high.

cs.SE↗

arXiv2025

Convergence rate of extreme eigenvalue of Ginibre ensembles to Gumbel distribution

Let $X$ be a real $(β=1)$ or complex $(β=2)$ Ginibre ensemble. Let $\{σ_i\}_{1\le i\le n}$ be the eigenvalues of $X,$ and $Z_n$ be some rescaled version of $\max_i \Re σ_i.$ It was proved that $Z_n$ converges weakly to the Gumbel distribution $Λ_β$ with distribution function $e^{-\fracβ{2}e^{-x}}.$ We further prove that $$\sup_{x\in \mathbb{R}}|\mathbb{P}(Z_n \leq x)-e^{-\fracβ{2}e^{-x}}|=\frac{25\log \log n}{4e \log n}(1+o(1))$$ and $$ W_1\left(\mathcal{L}(Z_n), Λ_β\right)=\frac{25\log \log n}{4\log n}(1+o(1))$$ for sufficiently large $n$, where $\mathcal{L}(Z_n)$ is the distribution of $Z_n$ and $W_1$ is the Wasserstein distance. Similar results hold for $\max_{i} |σ_i|.$ Furthermore, the convergence rates of the complex Ginibre ensemble are universal for complex iid random matrices under certain moment conditions on entries.

math.PR↗

arXiv2025

Revisit on the convergence rate of normal extremes

Let $(X_i)_{1 \le i \le n}$ be independent and identically distributed (i.i.d.) standard Gaussian random variables, and denote by $X_{(n)} = \max_{1 \le i \le n} X_i$ the maximum order statistic. It is well-known in extreme value theory that the linearly normalized maximum $ Y_n = a_n(X_{(n)} - b_n), $ converges weakly to the standard Gumbel distribution $\Lambda$ as $n \to \infty$, where $a_n > 0$ and $b_n$ are appropriate scaling and centering constants. In this note, choosing $$a_n=\sqrt{2\log n}\quad \text{and}\quad b_n = \sqrt{2 \log n} - \frac{\log \log n + \log (4\pi)}{2 \sqrt{2 \log n}},$$ we provide the exact order of this convergence under several distances including Berry-Esseen bound, $W_1$ distance, total variation distance, Kullback-Leibler divergence and Fisher information. We also show how the orders of these convergence are influenced by the choice of $b_n$ and $a_n.$

math.PR↗

arXiv2025

How fast does spectral radius of truncated circular unitary ensemble converge?

Let $z_1, \cdots, z_p$ be the eigenvalues of $A,$ which is the left-top $p\times p$ submatrix of an $n\times n$ Haar-invariant unitary matrix. Suppose there exist two constants $0<h_1<h_2<1$ such that $h_1<\frac pn<h_2.$ Then, $$\sup_{x\in \mathbb{R}}|\mathbb{P}(X_n\le x)-e^{-e^{-x}}|=\frac{(\log \log n)^{2}}{2e\log n}(1+o(1))$$ and further $$ W_{1}\left(\mathcal{L}(X_n),Λ\right)=\frac{(\log\log n)^2}{2\log n}(1+o(1))$$ for $n$ large enough. Here, $Λ$ is the Gumbel distribution and $\mathcal{L}(X_n)$ is the distribution of $X_n$ with $X_n$ being some rescaled version of $\max_{1\le i\le p}|z_i|,$ the spectral radius of $A.$

math.PR↗

arXiv2025

Universality of the convergence rate for spectral radius of complex IID random matrices

Let $X$ be an $n\times n$ matrix with independent and identically distributed entries $x_{ij} \stackrel{\text { d }}{=} n^{-1 / 2} x$ for some complex random variable $x$ of mean zero and variance one. Let $\{σ_i\}_{1\le i\le n}$ be the eigenvalues of $X$ and let $|σ_1|:=\max_{1\le i\le n}|σ_i|$ be the spectral radius. Set $Y_n=\sqrt{4 n γ_n}\left[|σ_1|-1-\sqrt{\frac{γ_n}{4 n}}\right],$ where $γ_{n}=\log{n}-2\log{\log{n}}-\log{2π}.$ As established in \cite{Cipolloni23Universality}, with specific moment-related conditions imposed on $x,$ the Gumbel distribution $Λ$ is identified as the universal weak limit of $Y_n.$ Subsequently, we extend this line of research and rigorously prove that the convergence rate, previously obtained for complex Ginibre ensembles in \cite{MaMeng25}, also possesses the property of universality. Precisely, one gets $$\sup_{x\in \mathbb{R}}|\mathbb{P}(Y_n \leq x)-e^{-e^{-x}}|=\frac{2\log\log n}{e\log n}(1+o(1))$$ and $$W_1\left(\mathcal{L}(Y_n), Λ\right)=\frac{2\log\log n}{\log n}(1+o(1))$$ for sufficiently large $n$, where $\mathcal{L}(Y_n)$ is the distribution of $Y_n$.

math.PR↗

arXiv2025

Fisher information approximation of random orthogonal matrices by Gaussian matrices

Let $Γ_n$ be an $n\times n$ Haar-invariant orthogonal matrix. Let ${ Z}_n$ be the $p\times q$ upper-left submatrix of $Γ_n$ and ${G}_n$ be a $p\times q$ matrix whose $pq$ entries are independent standard normals, where $p$ and $q$ are two positive integers. Let $\mathcal{L}(\sqrt{n} {Z}_n)$ and $\mathcal{L}({G}_n)$ be their joint distribution, respectively. Consider the Fisher information $I(\mathcal{L}(\sqrt{n} { Z}_n)|\mathcal{L}(G_n))$ between the distributions of $\sqrt{n} {Z}_n$ and ${ G}_n.$ In this paper, we conclude that $$I(\mathcal{L}(\sqrt{n} {Z}_n)|\mathcal{L}(G_n))\longrightarrow 0 $$ as $n\to\infty$ if $pq=o(n)$ and it does not tend to zero if $c=\lim\limits_{n\to\infty}\frac{pq}{n}\in(0, +\infty).$ Precisely, we obtain that $$I(\mathcal{L}(\sqrt{n} {Z}_n)|\mathcal{L}(G_n))=\frac{p^2q(q+1)}{4n^2}(1+o(1))$$ when $p=o(n).$

math.PR↗

arXiv2025

Optimal $W_1$ and Berry-Esseen bound between the spectral radius of large Chiral non-Hermitian random matrices and Gumbel

Consider the chiral non-Hermitian random matrix ensemble with parameters $n$ and $v$ and the non Hermiticity parameter $τ=0$ and let $(ζ_i)_{1\le i\le n}$ be its $n$ eigenvalues with positive $x$-coordinate. Set $$X_n:=\sqrt{\log s_n}\left(\frac{2n \max_{1\le i\le n}|ζ_i|^2-2\sqrt{n(n+v)}}{\sqrt{2n+v}}-a(s_{n})\right)$$ with $s_n=n(n+v)/(2n+v)$ and $a(s_n)=\sqrt{\log s_n}-\frac{\log(\sqrt{2π}\log s_n)}{\sqrt{\log s_n}}.$ It was proved in \cite{JQ} that $X_n$ converges weakly to the Gumbel distribution $Λ$. In this paper, we give in further that $$\lim_{n\to\infty} \frac{\log s_n}{(\log\log s_n)^2}W_1\left(F_n, Λ\right)=\frac{1}{2}$$ and the Berry-Esseen bound $$\lim_{n\to\infty} \frac{\log s_n}{(\log\log s_n)^2}\sup_{x\in\mathbb{R}}|F_n(x)-e^{-e^{-x}}|=\frac{1}{2e}.$$ Here, $F_n$ is the distribution (function) of $X_n.$

math.PR↗

arXiv2025

Exact convergence rate of spectral radius of complex Ginibre to Gumbel distribution

Consider the complex Ginibre ensemble, whose eigenvalues are $(λ_i)_{1\le i\le n}$ and the spectral radius $R_n=\max_{1\le i\le n}|λ_i|.$ Set $X_n=\sqrt{4 γ_{n}}(R_{n}-\sqrt{n}-\frac12\sqrt{γ_{n}})$ and $F_n$ be its distribution function, where $γ_{n}=\log n-2\log(\sqrt{2π}\log n).$ It was proved in \cite{Rider 2003} that $F_n$ converges weakly to the Gumbel distribution $Λ.$ We prove in further in this paper that $$\lim_{n\to\infty} \frac{\log n}{\log\log n}\, W_1\left(F_n, Λ\right)=2$$ and the Berry-Esseen bound $$\lim\limits_{n\to \infty} \frac{\log n}{\log\log n}\sup_{x\in \mathbb{R}}|F_{n}(x)-e^{-e^{-x}}|=\frac{2}{e}.$$

math.PR↗

arXiv2024

Deviation and moderate deviation for extremal eigenvalues of large Chiral non-Hermitian random matrices

Consider the chiral non-Hermitian random matrix ensemble with parameters $n$ and $v,$ and let $(ζ_i)_{1\le i\le n}$ be its $n$ eigenvalues with positive $x$-coordinate. In this paper, we establish deviation probabilities and moderate deviation probabilities for the spectral radius $(n/(n+v))^{1/2}\max_{1\le i\le n}|ζ_i|^2,$ as well as $(n/(n+v))^{1/2}\min_{1\le i\le n}|ζ_i|^2.$

math.PR↗