Searcharxiv⌕ Search

arXiv subjects

Xiaojin Zhang

Publications and source records attributed to Xiaojin Zhang.

At least 19 recordsLinked to original sources

The Auslander--Reiten Conjecture for Igusa--Todorov Algebras

We prove that every Igusa--Todorov Artin algebra satisfies the Auslander--Reiten conjecture. More precisely, let $V$ be an $n$-Igusa--Todorov witness, and let $t$ be the number of isomorphism classes of nonprojective indecomposable summands of $V$. If a finitely generated module $M$ satisfies $\Ext_A^i(M,A)=0$ for every $i>0$ and $\Ext_A^q(M,M)=0$ for $1\leq q\leq 2t+1$, then $M$ is projective. As applications, algebras of representation dimension at most three and algebras satisfying $J^{2m+1}=0$ for which $A/J^m$ has finite representation type satisfy the Auslander--Reiten conjecture.

math.RT↗

Comparing classical tilting modules and $τ$-tilting modules

We characterize support $τ$-tilting modules as classical tilting modules over quotient algebras satisfying a tensor-vanishing condition, and characterize classical tilting modules as $τ$-tilting modules satisfying a ${\rm Tor}^1$-vanishing condition. We use delooping levels to study \emph{Self-orthogonal $τ$-tilting Conjecture}: any self-orthogonal $τ$-tilting module is classical tilting. We confirm the conjecture when the endomorphism algebra of the module has finite global delooping level.

math.RT↗

Relative delooping levels on intervals of torsion classes

Let $U$ be a support $τ$-tilting module over a finite-dimensional algebra $A$, and let $\mathcal{V}\subseteq\operatorname{Fac}U$ be any torsion class. The exact heart $\mathcal{H}=\operatorname{Fac}U\cap\mathcal{V}^{\perp}$ has an Ext-progenerator $P$. We prove that $\operatorname{Hom}_A(P,-)$ realizes $\mathcal{H}$ as a submodule-closed exact subcategory of $\operatorname{mod}B$, where $B=\text{End}_A(P)$, and preserves minimal syzygies. The global relative $k$-delooping levels are independent of $k\geq1$. Writing $g=\operatorname{glo}\text{-}\operatorname{dell}\mathcal{H}$, we obtain the sharp comparison $g\leq\operatorname{glo}\text{-}\operatorname{dell}B\leq g+1$ and $\operatorname{findim}B^{\mathrm{op}}\leq g+1$. Thus the two global levels are finite simultaneously. For wide hearts the relative levels equal their classical counterparts. We also characterize syzygy finiteness of $B$ by a finite relative syzygy stratum and obtain explicit bounds for both finitistic dimensions. If the heart is Frobenius, it is exactly the category of finitely generated Gorenstein-projective $B$-modules under this realization, and both self-injective dimensions of $B$ are at most one. We also give examples to show the main results.

math.RT↗

FedPix: Learning What to Share with Less Leakage in Federated Intelligence

Existing federated intelligence paradigms primarily focus on how models are trained or aggregated, while overlooking a more fundamental question: \textbf{what information should leave the client?} Existing methods often share model updates that users cannot easily inspect, or data representations that may reveal sensitive information. Moreover, generating these representations can be computationally expensive for resource-constrained clients. We observe that image classification does not require all the visual details needed to reconstruct the original image. This suggests that clients could share less information while preserving task performance. However, learning what to preserve is challenging when the server cannot access raw client data. To address this challenge, we propose \textbf{FedPix}, which learns to share compact, visually inspectable images instead of model updates or intermediate features. The key idea is to use the shared representations themselves to learn what information should be preserved and what can be discarded. By balancing task utility and reconstruction resistance, FedPix enables the server to improve the representations without accessing raw client data, while keeping client-side computation lightweight. Extensive experiments on BloodMNIST, CIFAR-10, FashionMNIST, and MNIST demonstrate that FedPix achieves strong reconstruction resistance, with reconstruction PSNR as low as 6.87~dB and synthetic PSNR as low as 6.74~dB. Meanwhile, FedPix reduces client runtime by up to 96.0\% (from 1223~s to 49~s on CIFAR-10), while maintaining competitive classification accuracy of 99.18\% on MNIST, 89.97\% on FashionMNIST, 82.24\% on BloodMNIST, and 65.63\% on CIFAR-10.

cs.CR↗

Variance-Aware Non-Stationary Bandits:All-Budget Regret Bounds and Adaptation Costs

Low reward variance reduces estimation cost, but does not eliminate monitoring of arms that may improve while unobserved. We refine the variance--variation tradeoff and characterize how both costs depend on supplied variance and variation budgets in bounded, non-stationary multi-armed bandits. Let $A$ be the number of arms, $T$ the horizon, $B$ a bound on consecutive mean variation excluding the initial vector, and $Q$ a bound on the sum of the largest per-round reward variances. For every admissible pair $Q,B$, the minimax expected dynamic regret is, within a single logarithmic factor, \[ \min\{T,\;A+\sqrt{AQ}+(AQBT)^{1/3} +\sqrt{AT}\min\{B,\sqrt B\}\}. \] The deterministic boundary $Q=0$ is characterized without logarithms. For small positive budgets, a safe calibration phase identifies a common interval containing all future means of surviving arms. Clipping observations to a wider interval reduces the cost of restarting a prior hybrid static-regret algorithm; an original-mean analytical reference controls its residuals without charging the variation of clipped means. Balancing clipping bias and monitoring closes the small-budget noisy regime. Fixed oblivious constructions give the matching lower bound. We also quantify adaptation costs: on a disclosed deterministic one-improvement family, a stationary-regret budget has a matching worst-case change-regret frontier, including a finite improvement pulse. This implies an $Ω((T/A)^{1/4})$ loss when one policy competes with all supplied-budget deterministic oracle rates. The budget-free low-noise scheduling problem remains separate.

cs.LG↗

Mitigating Data Reconstruction Attacks in Federated Learning: Conditions, Bounds, and Theoretical Guarantees

Privacy attacks, particularly data reconstruction attacks, pose significant challenges to federated learning despite its promise in privacy-preserving collaborative training. This paper presents a comprehensive theoretical analysis of privacy leakage in federated learning through dual perspectives: linear algebra and optimization theory. By analyzing the client-update Jacobian, we identify conditions under which distinct data batches produce the same model update and derive a sufficient batch-size criterion for non-unique reconstruction. Through optimization theory, we establish a novel upper bound for privacy leakage that incorporates batch size, distortion extent, and other critical parameters. These theoretical insights elucidate the fundamental relationships between system parameters and privacy vulnerability, laying the groundwork for developing more robust privacy-preserving federated learning frameworks.

cs.CR↗

FUSE: Fairness-aware Update Screening against Byzantine Exploitation in $q$-Fair Federated Learning

Client-level fair federated learning, in the form of $q$-fair federated learning ($q$-FFL), agnostic federated learning, or tilted empirical risk minimization, has emerged as an attractive solution to equalize utility across heterogeneous clients. This paradigm grants high-loss clients greater relative influence, while a central server aggregates their loss-weighted updates into a single shared model. Yet, the Byzantine robustness of $q$-fair federated learning has been largely neglected, leaving the server unable to tell legitimate fairness-induced amplification from malicious update scaling. To address this critical gap, we formalize the problem of Byzantine-robust $q$-fair federated learning under an \emph{update-only interface}, in which the server observes nothing but the fairness-scaled client updates. Our analysis under shared class-conditional label skew shows that fairness scaling preserves the nonnegative class-mixture structure of honest updates, while replacing their unit-sum constraint by a client-dependent radial scale. Moving beyond magnitude-based screening, we show that this warped geometry supplies two complementary tests, one radial and one structural, that a single norm cannot provide. We then introduce FUSE, a structure-aware robust aggregator that uses a small trusted labeled dataset for coefficient recovery, scale and cone-consistency screening, and post-screening step-size control. Across four benchmarks and seven Byzantine attacks, FUSE improves seed-wise worst-attack accuracy over the strongest baseline on each benchmark by 3.9--30.1 percentage points, while largely preserving the influence of the high-loss honest clients that fairness is meant to protect.

cs.LG↗

Lost in Implicit Routing: How LLM Agents Fail to Load Relevant Skills at Scale

Reusable skills let LLM agents load task-specific instructions on demand, but they make execution depend on implicit routing over skill metadata. We test whether this design scales by holding SkillsBench tasks and their golden skills fixed while adding irrelevant skills. We find skill routing collapse: agents increasingly load fewer or no skills, even though the relevant skills remain available. With 1,000 irrelevant skills, loaded-skill recall falls from 73.6% to 27.4% and the zero-load rate rises from 9.3% to 51.6%, whereas explicit selection from the same library retains 73.9% recall. Loading other skills, including plausible alternatives to golden skills, is too rare to explain this decline. Masking only the descriptions of irrelevant skills largely restores skill loading, yet with the metadata intact, explicitly asking agents to plan skill use still yields lower recall than preselection. These findings point to semantic interference from descriptions and instruction interference when skill choice competes with task execution. We also evaluate explicit preselection as a practical mitigation, including its coverage and latency trade-offs. Our analysis informs skill protocols and evaluation for agents with growing skill libraries.

cs.AI↗

Near-Linear Regimes for One-Sided Distribution-Free Junta Testing

We study adaptive one-sided distribution-free testing of Boolean $k$-juntas, with membership-query access to the function and sample access to an arbitrary unknown distribution. For general Boolean functions, we give a tester using $O(k^2\log(k+1)+k/ε)$ queries, independently of the ambient dimension. The tester maintains disjoint relevant blocks certified by observed disagreements. Within each fresh batch, it jointly decodes literal signatures for a small family of random base masks, then uses linear combinations to generate $Θ(1/ε)$ probes. Their ideal counterparts are pairwise independent, yielding constant conditional progress probability; a global $O(k)$ batch budget gives the additive query bound. Consequently, the complexity is $O((k/ε)\log(k+1))$ when $ε\le1/k$, and $O(k/ε)$ when $ε\le1/(k\log(k+1))$. We also analyze an indexed-table implementation that reuses previously computed block orientations. For functions symmetric on an unknown set of relevant variables, it uses $O(k\log^2(k+1)/ε)$ queries: safely checked blocks contain a single globally relevant variable, making their cached nonliteral sides reusable. For arbitrary functions, a stable-deletion analysis establishes soundness of the same implementation at $O(k^2\log(k+1)/ε)$ queries. All variants always accept $k$-juntas, and every rejection is certified by $k+1$ disjoint relevant blocks. The fresh-batch construction and the reuse analysis use the same certified-block framework, with different randomness and stopping rules.

cs.CC↗

Relative Suspension and Higher Delooping for Support $τ$-Tilting Modules

We construct relative suspension for the exact category generated by a support $τ$-tilting module and characterize higher delooping by the units of the resulting adjunction, which are the generalizations of Gélinas's results in Adv. Math.(2022). We prove that comparison with ordinary higher delooping requires a shift of one in the index. More precisely, for every $k,d\ge1$, there is a module whose ordinary $k$-delooping level is zero and whose relative level is $d$. We also prove that tensor induction preserves the relative constructions and higher delooping levels on induced modules. As an application, we study cyclic Nakayama algebras with arbitrary local coefficients and automorphism twists. Their left and right little and big finitistic dimensions and ordinary and derived delooping levels all equal the finitistic dimension of the underlying Nakayama algebra.

math.RT↗

Self-Orthogonal $τ$-Tilting Modules and Tilting Modules III: Finitistic Dimension

Let $A$ be an Artin algebra and let $T$ be a $τ$-tilting right $A$-module. Put $C=A/{\rm Ann}_A(T)$. We prove that finite little finitistic dimension of $C^{\rm op}$, together with a finite range of self-extension vanishing for $T$, forces $T$ to be a classical $1$-tilting module. We also show that if the finitistic conjecture holds for all Artin algebras then the self-orthogonal $τ$-tilting conjecture holds for all Artin algebras. Moreover, a counterexample to the self-orthogonal $τ$-tilting conjecture will give a counterexample to the finitistic dimension conjecture. As an application, the self-orthogonal $τ$-tilting conjecture holds for Artin algebras with radical cube zero.

math.RT↗

The Auslander-Reiten conjecture for algebras with radical cube zero

Let $A$ be a split finite-dimensional algebra over a field whose radical $J$ satisfies $J^3=0$, and let $s$ be the number of isomorphism classes of simple $A$-modules. We prove that a non-projective module $M$ with ${\rm Ext}_A^i(M,A)=0$ for all $i>0$ has a non-zero self-extension in some degree between $1$ and $3s+1$. In particular, $A$ satisfies the Auslander--Reiten conjecture, which asserts that every self-orthogonal generator is projective. As a consequence, every finite-dimensional algebra over an algebraically closed field with radical cube zero satisfies the Auslander-Reiten conjecture.

math.RT↗

When Evidence Shapes Collaboration: Knowledge-Conditioned Topology Generation for Multi-Agent Systems

Multi-Agent Systems (MAS) have recently moved from static workflows toward dynamically generated collaboration topologies. However, existing topology generation methods rely primarily on the parametric knowledge of large language models, with external search or retrieval used only as a reactive tool rather than an explicit determinant of collaboration structure. This leads to structure-knowledge misalignment, where systems exhibit redundant interactions or insufficient verification in knowledge-intensive tasks. We propose K-GAT (Knowledge-Guided Agent Topology Generator), a neuro-symbolic framework that formulates collaboration topology design as a knowledge-conditioned structure learning problem, integrating external evidence directly into autoregressive graph generation. Extensive experiments on knowledge-intensive benchmarks demonstrate K-GAT's efficiency and effectiveness: notably on the expert-level GPQA dataset, K-GAT outperforms the LLM-Debate baseline by a substantial margin of +15.7% in accuracy, while consuming less than half the computational tokens.

cs.AI↗

RIDGE: Region-Informed Derivative-Guided Evidence Selection for Long Video Understanding

Long videos contain far more visual content than Large Vision-Language Models (LVLMs) can process under a fixed visual-token budget, making frame selection essential. Existing query-aware selectors usually estimate frame-query relevance and build a compact subset from high-scoring frames. Although their mechanisms differ, the similarity sequence is still often treated primarily as values to rank or sample from, rather than as an ordered signal whose shape reflects how query-relevant evidence emerges, peaks, and fades over time. This can obscure frames that explain, contextualize, or follow an event, because such evidence may lie on the rising or falling sides of a nearby relevance peak and receive lower absolute scores. We propose RIDGE, a frame selection framework that reads the frame-query similarity curve as a temporal signal. By using local changes and curvature, RIDGE partitions the timeline into structural regions and applies region-specific selection to preserve event cores, transitions, buildup, aftermath, and contextual frames under a fixed budget. It is a lightweight post-processing step on precomputed frame-query scores and requires neither training nor iterative LVLM calls. Across four long-video benchmarks and three backbones, RIDGE achieves the best performance in most settings and remains competitive in the others.

cs.CV↗

Self-Orthogonal $τ$-Tilting Modules and Tilting Modules II: Annihilator Separation

Let $A$ be an Artin algebra and let $T$ be a self-orthogonal $τ$-tilting right $A$-module. Set $I=\Ann_A(T)$. We prove that \[ \Hom_A(I,T)=0=\Ext_A^n(I,T)\qquad(n\geq 1). \] It follows that the torsion class generated by $I$ is Hom-orthogonal to $\Sub T$. This separation yields faithfulness criteria expressed through ideal tops, socles, and projective supports. We also prove a finitistic-dimension obstruction: if $B=\End_A(T)^{\rm op}$ has finite little finitistic dimension, then $\Fac T\cap{}^{\perp_{\geq0}}T=\{0\}$. As an application, the self-orthogonal $τ$-tilting conjecture holds for radical square zero Artin algebras. Finally, the support criteria apply to semisimple-source triangular matrix algebras with arbitrary local terminal blocks.

math.RT↗

Theoretically Principled Federated Learning for Balancing Privacy and Utility

We propose a general learning framework for the protection mechanisms that protects privacy via distorting model parameters, which facilitates the trade-off between privacy and utility. The algorithm is applicable to arbitrary privacy measurements that maps from the distortion to a real value. It can achieve personalized utility-privacy trade-off for each model parameter, on each client, at each communication round in federated learning. Such adaptive and fine-grained protection can improve the effectiveness of privacy-preserved federated learning. Theoretically, we show that gap between the utility loss of the protection hyperparameter output by our algorithm and that of the optimal protection hyperparameter is sub-linear in the total number of iterations. The sublinearity of our algorithm indicates that the average gap between the performance of our algorithm and that of the optimal performance goes to zero when the number of iterations goes to infinity. Further, we provide the convergence rate of our proposed algorithm. We conduct empirical results on benchmark datasets to verify that our method achieves better utility than the baseline methods under the same privacy budget.

cs.LG↗

MedReaMM: Evaluating Large Multimodal Models on Expert-Level Clinical Diagnostic Synthesis

The application of Large Language Models (LLMs) to diagnostic decision-making has garnered growing interest. However, existing benchmarks largely focus on textual reasoning or isolated visual question-answering (VQA) tasks, lacking holistic integration of clinical narratives and medical imaging, and thus failing to assess the multimodal diagnostic synthesis capability central to expert clinical judgment. To bridge this gap, we introduce MedReaMM, a benchmark specifically designed to evaluate models' ability to synthesize heterogeneous clinical evidence consisting of detailed patient histories alongside multiple medical images into accurate differential diagnoses under a complete-information paradigm. Constructed from case reports sourced from top-tier medical journals and curated clinical case databases, MedReaMM comprises 625 expert-validated cases with an average of 2.79 medical images per case and a total of 1,042 standardized diagnoses annotated with ICD-11 codes. These cases predominantly represent rare, atypical, or multi-system presentations that demand expert-level evidence integration beyond routine pattern recognition. We evaluate 23 Large Multimodal Models (LMMs) and find that most achieve diagnostic accuracy scores below 50%, underscoring a substantial gap in multimodal diagnostic synthesis capability. Further analysis reveals that medical knowledge proficiency, medical image understanding, and evidence integration are all highly correlated with diagnostic performance.

cs.CV↗

Doc-V*:Coarse-to-Fine Interactive Visual Reasoning for Multi-Page Document VQA

Multi-page Document Visual Question Answering requires reasoning over semantics, layouts, and visual elements in long, visually dense documents. Existing OCR-free methods face a trade-off between capacity and precision: end-to-end models scale poorly with document length, while visual retrieval-based pipelines are brittle and passive. We propose Doc-$V^*$, an \textbf{OCR-free agentic} framework that casts multi-page DocVQA as sequential evidence aggregation. Doc-$V^*$ begins with a thumbnail overview, then actively navigates via semantic retrieval and targeted page fetching, and aggregates evidence in a structured working memory for grounded reasoning. Trained by imitation learning from expert trajectories and further optimized with Group Relative Policy Optimization, Doc-$V^*$ balances answer accuracy with evidence-seeking efficiency. Across five benchmarks, Doc-$V^*$ outperforms open-source baselines and approaches proprietary models, improving out-of-domain performance by up to \textbf{47.9\%} over RAG baseline. Other results reveal effective evidence aggregation with selective attention, not increased input pages.

cs.CL↗