arXiv subjects

Xiangyu Chang

Publications and source records attributed to Xiangyu Chang.

At least 19 recordsLinked to original sources

arXiv2026

Fairness May Backfire: When Leveling-Down Occurs in Fair Machine Learning

As machine learning (ML) systems increasingly shape access to credit, jobs, and other opportunities, the fairness of algorithmic decisions has become a central concern. Yet it remains unclear when enforcing fairness constraints in these systems genuinely improves outcomes for affected groups or instead leads to "leveling down," making one or both groups worse off. We address this question in a unified, population-level (Bayes) framework for binary classification under prevalent group fairness notions. Our Bayes approach is distribution-free and algorithm-agnostic, isolating the intrinsic effect of fairness requirements from finite-sample noise and from training and intervention specifics. We analyze two deployment regimes for ML classifiers under common legal and governance constraints: attribute-aware decision-making (sensitive attributes available at decision time) and attribute-blind decision-making (sensitive attributes excluded from prediction). We show that, in the attribute-aware regime, fair ML necessarily (weakly) improves outcomes for the disadvantaged group and (weakly) worsens outcomes for the advantaged group. In contrast, in the attribute-blind regime, the impact of fairness is distribution-dependent: fairness can benefit or harm either group and may shift both groups' outcomes in the same direction, leading to either leveling up or leveling down. We characterize the conditions under which these patterns arise and highlight the role of "masked" candidates in driving them. Overall, our results provide structural guidance on when pursuing algorithmic fairness is likely to improve group outcomes and when it risks systemic leveling down, informing fair ML design and deployment choices.

stat.ML↗

arXiv2026

Beyond Cross-Validation: Adaptive Parameter Selection for Kernel-Based Gradient Descents

This paper proposes a novel parameter selection strategy for kernel-based gradient descent (KGD) algorithms, integrating bias-variance analysis with the splitting method. We introduce the concept of empirical effective dimension to quantify iteration increments in KGD, deriving an adaptive parameter selection strategy that is implementable. Theoretical verifications are provided within the framework of learning theory. Utilizing the recently developed integral operator approach, we rigorously demonstrate that KGD, equipped with the proposed adaptive parameter selection strategy, achieves the optimal generalization error bound and adapts effectively to different kernels, target functions, and error metrics. Consequently, this strategy showcases significant advantages over existing parameter selection methods for KGD.

stat.ML↗

arXiv2025

AKG kernel Agent: A Multi-Agent Framework for Cross-Platform Kernel Synthesis

Modern AI models demand high-performance computation kernels. The growing complexity of LLMs, multimodal architectures, and recommendation systems, combined with techniques like sparsity and quantization, creates significant computational challenges. Moreover, frequent hardware updates and diverse chip architectures further complicate this landscape, requiring tailored kernel implementations for each platform. However, manual optimization cannot keep pace with these demands, creating a critical bottleneck in AI system development. Recent advances in LLM code generation capabilities have opened new possibilities for automating kernel development. In this work, we propose AKG kernel agent (AI-driven Kernel Generator), a multi-agent system that automates kernel generation, migration, and performance tuning. AKG kernel agent is designed to support multiple domain-specific languages (DSLs), including Triton, TileLang, CPP, and CUDA-C, enabling it to target different hardware backends while maintaining correctness and portability. The system's modular design allows rapid integration of new DSLs and hardware targets. When evaluated on KernelBench using Triton DSL across GPU and NPU backends, AKG kernel agent achieves an average speedup of 1.46$\times$ over PyTorch Eager baselines implementations, demonstrating its effectiveness in accelerating kernel development for modern AI workloads.

cs.AI↗

arXiv2025

Mitigating Participation Imbalance Bias in Asynchronous Federated Learning

In Asynchronous Federated Learning (AFL), the central server immediately updates the global model with each arriving client's contribution. As a result, clients perform their local training on different model versions, causing information staleness (delay). In federated environments with non-IID local data distributions, this asynchronous pattern amplifies the adverse effect of client heterogeneity (due to different data distribution, local objectives, etc.), as faster clients contribute more frequent updates, biasing the global model. We term this phenomenon heterogeneity amplification. Our work provides a theoretical analysis that maps AFL design choices to their resulting error sources when heterogeneity amplification occurs. Guided by our analysis, we propose ACE (All-Client Engagement AFL), which mitigates participation imbalance through immediate, non-buffered updates that use the latest information available from all clients. We also introduce a delay-aware variant, ACED, to balance client diversity against update staleness. Experiments on different models for different tasks across diverse heterogeneity and delay settings validate our analysis and demonstrate the robust performance of our approaches.

cs.LG↗

arXiv2025

Rethinking Data Value: Asymmetric Data Shapley for Structure-Aware Valuation in Data Markets and Machine Learning Pipelines

Rigorous valuation of individual data sources is critical for fair compensation in data markets, informed data acquisition, and transparent development of ML/AI models. Classical Data Shapley (DS) provides a essential axiomatic framework for data valuation but is constrained by its symmetry axiom that assumes interchangeability of data sources. This assumption fails to capture the directional and temporal dependencies prevalent in modern ML/AI workflows, including the reliance of duplicated or augmented data on original sources and the order-specific contributions in sequential pipelines such as federated learning and multi-stage LLM fine tuning. To address these limitations, we introduce Asymmetric Data Shapley (ADS), a structure-aware data valuation framework for modern ML/AI pipelines. ADS relaxes symmetry by averaging marginal contributions only over permutations consistent with an application-specific ordering of data groups. It preserves efficiency and linearity, maintains within group symmetry and directional precedence across groups, and reduces to DS when the ordering collapses to a single group. We develop two complementary computational procedures for ADS: (i) a Monte Carlo estimator (MC-ADS) with finite-sample accuracy guarantees, and (ii) a k-nearest neighbor surrogate (KNN-ADS) that is exact and efficient for KNN predictors. Across representative settings with directional and temporal dependence, ADS consistently outperforms benchmark methods by distinguishing novel from redundant contributions and respecting the sequential nature of training. These results establish ADS as a principled and practical approach to equitable data valuation in data markets and complex ML/AI pipelines.

cs.GT↗

arXiv2025

Beyond Uniform Deletion: A Data Value-Weighted Framework for Certified Machine Unlearning

As the right to be forgotten becomes legislated worldwide, machine unlearning mechanisms have emerged to efficiently update models for data deletion and enhance user privacy protection. However, existing machine unlearning algorithms frequently neglect the fact that different data points may contribute unequally to model performance (i.e., heterogeneous data values). Treat them equally in machine unlearning procedure can potentially degrading the performance of updated models. To address this limitation, we propose Data Value-Weighted Unlearning (DVWU), a general unlearning framework that accounts for data value heterogeneity into the unlearning process. Specifically, we design a weighting strategy based on data values, which are then integrated into the unlearning procedure to enable differentiated unlearning for data points with varying utility to the model. The DVWU framework can be broadly adapted to various existing machine unlearning methods. We use the one-step Newton update as an example for implementation, developing both output and objective perturbation algorithms to achieve certified unlearning. Experiments on both synthetic and real-world datasets demonstrate that our methods achieve superior predictive performance and robustness compared to conventional unlearning approaches. We further show the extensibility of our framework on gradient ascent method by incorporating the proposed weighting strategy into the gradient terms, highlighting the adaptability of DVWU for broader gradient-based deep unlearning methods.

cs.LG↗

arXiv2025

VDSAgents: A PCS-Guided Multi-Agent System for Veridical Data Science Automation

Large language models (LLMs) become increasingly integrated into data science workflows for automated system design. However, these LLM-driven data science systems rely solely on the internal reasoning of LLMs, lacking guidance from scientific and theoretical principles. This limits their trustworthiness and robustness, especially when dealing with noisy and complex real-world datasets. This paper provides VDSAgents, a multi-agent system grounded in the Predictability-Computability-Stability (PCS) principles proposed in the Veridical Data Science (VDS) framework. Guided by PCS principles, the system implements a modular workflow for data cleaning, feature engineering, modeling, and evaluation. Each phase is handled by an elegant agent, incorporating perturbation analysis, unit testing, and model validation to ensure both functionality and scientific auditability. We evaluate VDSAgents on nine datasets with diverse characteristics, comparing it with state-of-the-art end-to-end data science systems, such as AutoKaggle and DataInterpreter, using DeepSeek-V3 and GPT-4o as backends. VDSAgents consistently outperforms the results of AutoKaggle and DataInterpreter, which validates the feasibility of embedding PCS principles into LLM-driven data science automation.

cs.AI↗

arXiv2025

A Unified Zeroth-Order Optimization Framework via Oblivious Randomized Sketching

We propose a new framework for analyzing zeroth-order optimization (ZOO) from the perspective of \emph{oblivious randomized sketching}.In this framework, commonly used gradient estimators in ZOO-such as finite difference (FD) and random finite difference (RFD)-are unified through a general sketch-based formulation. By introducing the concept of oblivious randomized sketching, we show that properly chosen sketch matrices can significantly reduce the high variance of RFD estimates and enable \emph{high-probability} convergence guarantees of ZOO, which are rarely available in existing RFD analyses. \noindent We instantiate the framework on convex quadratic objectives and derive a query complexity of $\tilde{\mathcal{O}}(\mathrm{tr}(A)/L \cdot L/\mu\log\frac{1}{\epsilon})$ to achieve a $\epsilon$-suboptimal solution, where $A$ is the Hessian, $L$ is the largest eigenvalue of $A$, and $\mu$ denotes the strong convexity parameter. This complexity can be substantially smaller than the standard query complexity of ${\cO}(d\cdot L/\mu \log\frac{1}{\epsilon})$ that is linearly dependent on problem dimensionality, especially when $A$ has rapidly decaying eigenvalues. These advantages naturally extend to more general settings, including strongly convex and Hessian-aware optimization. \noindent Overall, this work offers a novel sketch-based perspective on ZOO that explains why and when RFD-type methods can achieve \emph{weakly dimension-independent} convergence in general smooth problems, providing both theoretical foundations and practical implications for ZOO.

math.OC↗

arXiv2025

Can a One-Point Feedback Zeroth-order Algorithm Achieve Linear Dimension Dependent Sample Complexity?

We revisit the one-point feedback zeroth-order (ZO) optimization problem, a classical setting in derivative-free optimization where only a single noisy function evaluation is available per query. Compared to their two-point counterparts, existing one-point feedback ZO algorithms typically suffer from poor dimension dependence in their sample complexities -- often quadratic or worse -- even for convex problems. This gap has led to the open question of whether one-point feedback ZO algorithms can match the optimal \emph{linear} dimension dependence achieved by two-point methods. In this work, we answer this question \emph{affirmatively}.

math.OC↗

arXiv2025

When and How Unlabeled Data Provably Improve In-Context Learning

Recent research shows that in-context learning (ICL) can be effective even when demonstrations have missing or incorrect labels. To shed light on this capability, we examine a canonical setting where the demonstrations are drawn according to a binary Gaussian mixture model (GMM) and a certain fraction of the demonstrations have missing labels. We provide a comprehensive theoretical study to show that: (1) The loss landscape of one-layer linear attention models recover the optimal fully-supervised estimator but completely fail to exploit unlabeled data; (2) In contrast, multilayer or looped transformers can effectively leverage unlabeled data by implicitly constructing estimators of the form $\sum_{i\ge 0} a_i (X^\top X)^iX^\top y$ with $X$ and $y$ denoting features and partially-observed labels (with missing entries set to zero). We characterize the class of polynomials that can be expressed as a function of depth and draw connections to Expectation Maximization, an iterative pseudo-labeling algorithm commonly used in semi-supervised learning. Importantly, the leading polynomial power is exponential in depth, so mild amount of depth/looping suffices. As an application of theory, we propose looping off-the-shelf tabular foundation models to enhance their semi-supervision capabilities. Extensive evaluations on real-world datasets show that our method significantly improves the semisupervised tabular learning performance over the standard single pass inference.

cs.LG↗

arXiv2025

Stochastic Diagonal Estimation Based on Matrix Quadratic Form Oracles

We study the problem of estimating the diagonal of an implicitly given matrix $\Ab$. For such a matrix we have access to an oracle that allows us to evaluate the matrix quadratic form $ \ub^\top \Ab \ub$. Based on this query oracle, we propose a stochastic diagonal estimation method with random variable $\ub$ drawn from the standard Gaussian distribution. We provide the element-wise and norm-wise sample complexities of the proposed method. Our numerical experiments on different types and dimensions matrices demonstrate the effectiveness of our method and validate the tightness of theoretical results.

math.NA↗

arXiv2025

Privacy Leaks by Adversaries: Adversarial Iterations for Membership Inference Attack

Membership inference attack (MIA) has become one of the most widely used and effective methods for evaluating the privacy risks of machine learning models. These attacks aim to determine whether a specific sample is part of the model's training set by analyzing the model's output. While traditional membership inference attacks focus on leveraging the model's posterior output, such as confidence on the target sample, we propose IMIA, a novel attack strategy that utilizes the process of generating adversarial samples to infer membership. We propose to infer the member properties of the target sample using the number of iterations required to generate its adversarial sample. We conduct experiments across multiple models and datasets, and our results demonstrate that the number of iterations for generating an adversarial sample is a reliable feature for membership inference, achieving strong performance both in black-box and white-box attack scenarios. This work provides a new perspective for evaluating model privacy and highlights the potential of adversarial example-based features for privacy leakage assessment.

cs.CR↗

arXiv2025

Continuum-armed Bandit Optimization with Batch Pairwise Comparison Oracles

This paper studies a bandit optimization problem where the goal is to maximize a function $f(x)$ over $T$ periods for some unknown strongly concave function $f$. We consider a new pairwise comparison oracle, where the decision-maker chooses a pair of actions $(x, x')$ for a consecutive number of periods and then obtains an estimate of $f(x)-f(x')$. We show that such a pairwise comparison oracle finds important applications to joint pricing and inventory replenishment problems and network revenue management. The challenge in this bandit optimization is twofold. First, the decision-maker not only needs to determine a pair of actions $(x, x')$ but also a stopping time $n$ (i.e., the number of queries based on $(x, x')$). Second, motivated by our inventory application, the estimate of the difference $f(x)-f(x')$ is biased, which is different from existing oracles in stochastic optimization literature. To address these challenges, we first introduce a discretization technique and local polynomial approximation to relate this problem to linear bandits. Then we developed a tournament successive elimination technique to localize the discretized cell and run an interactive batched version of LinUCB algorithm on cells. We establish regret bounds that are optimal up to poly-logarithmic factors. Furthermore, we apply our proposed algorithm and analytical framework to the two operations management problems and obtain results that improve state-of-the-art results in the existing literature.

cs.LG↗

arXiv2025

Bayes-Optimal Fair Classification with Multiple Sensitive Features

Existing theoretical work on Bayes-optimal fair classifiers usually considers a single (binary) sensitive feature. In practice, individuals are often defined by multiple sensitive features. In this paper, we characterize the Bayes-optimal fair classifier for multiple sensitive features under general approximate fairness measures, including mean difference and mean ratio. We show that these approximate measures for existing group fairness notions, including Demographic Parity, Equal Opportunity, Predictive Equality, and Accuracy Parity, are linear transformations of selection rates for specific groups defined by both labels and sensitive features. We then characterize that Bayes-optimal fair classifiers for multiple sensitive features become instance-dependent thresholding rules that rely on a weighted sum of these group membership probabilities. Our framework applies to both attribute-aware and attribute-blind settings and can accommodate composite fairness notions like Equalized Odds. Building on this, we propose two practical algorithms for Bayes-optimal fair classification via in-processing and post-processing. We show empirically that our methods compare favorably to existing methods.

stat.ML↗

arXiv2025

Breaking the Prompt Wall (I): A Real-World Case Study of Attacking ChatGPT via Lightweight Prompt Injection

This report presents a real-world case study demonstrating how prompt injection can attack large language model platforms such as ChatGPT according to a proposed injection framework. By providing three real-world examples, we show how adversarial prompts can be injected via user inputs, web-based retrieval, and system-level agent instructions. These attacks, though lightweight and low-cost, can cause persistent and misleading behaviors in LLM outputs. Our case study reveals that even commercial-grade LLMs remain vulnerable to subtle manipulations that bypass safety filters and influence user decisions. \textbf{More importantly, we stress that this report is not intended as an attack guide, but as a technical alert. As ethical researchers, we aim to raise awareness and call upon developers, especially those at OpenAI, to treat prompt-level security as a critical design priority.

cs.CR↗

arXiv2025

Privacy-Preserving Transfer Learning for Community Detection using Locally Distributed Multiple Networks

Modern applications increasingly involve highly sensitive network data, where raw edges cannot be shared due to privacy constraints. We propose \texttt{TransNet}, a new spectral clustering-based transfer learning framework that improves community detection on a \emph{target network} by leveraging heterogeneous, locally stored, and privacy-preserved auxiliary \emph{source networks}. Our focus is the \textit{local differential privacy} regime, in which each local data provider perturbs edges via \textit{randomized response} before release, requiring no trusted third party. \texttt{TransNet} aggregates source eigenspaces through a novel adaptive weighting scheme that accounts for both privacy and heterogeneity, and then regularizes the weighted source eigenspace with the target eigenspace to optimally balance the two. Theoretically, we establish an error-bound-oracle property: the estimation error for the aggregated eigenspace depends only on \textit{informative sources}, ensuring robustness when some sources are highly heterogeneous or heavily privatized. We further show that the error bound of \texttt{TransNet} is no greater than that of estimators using only the target network or only (weighted) sources. Empirically, \texttt{TransNet} delivers strong gains across a range of privacy levels and heterogeneity patterns. For completeness, we also present \texttt{TransNetX}, an extension based on Gaussian perturbation of projection matrices under the assumption that trusted local data curators are available.

stat.ML↗

arXiv2025

Provable Benefits of Task-Specific Prompts for In-context Learning

The in-context learning capabilities of modern language models have motivated a deeper mathematical understanding of sequence models. A line of recent work has shown that linear attention models can emulate projected gradient descent iterations to implicitly learn the task vector from the data provided in the context window. In this work, we consider a novel setting where the global task distribution can be partitioned into a union of conditional task distributions. We then examine the use of task-specific prompts and prediction heads for learning the prior information associated with the conditional task distribution using a one-layer attention model. Our results on loss landscape show that task-specific prompts facilitate a covariance-mean decoupling where prompt-tuning explains the conditional mean of the distribution whereas the variance is learned/explained through in-context learning. Incorporating task-specific head further aids this process by entirely decoupling estimation of mean and variance components. This covariance-mean perspective similarly explains how jointly training prompt and attention weights can provably help over fine-tuning after pretraining.

cs.CL↗

arXiv2025

An Enhanced Zeroth-Order Stochastic Frank-Wolfe Framework for Constrained Finite-Sum Optimization

We propose an enhanced zeroth-order stochastic Frank-Wolfe framework to address constrained finite-sum optimization problems, a structure prevalent in large-scale machine-learning applications. Our method introduces a novel double variance reduction framework that effectively reduces the gradient approximation variance induced by zeroth-order oracles and the stochastic sampling variance from finite-sum objectives. By leveraging this framework, our algorithm achieves significant improvements in query efficiency, making it particularly well-suited for high-dimensional optimization tasks. Specifically, for convex objectives, the algorithm achieves a query complexity of O(d \sqrt{n}/\epsilon ) to find an epsilon-suboptimal solution, where d is the dimensionality and n is the number of functions in the finite-sum objective. For non-convex objectives, it achieves a query complexity of O(d^{3/2}\sqrt{n}/\epsilon^2 ) without requiring the computation ofd partial derivatives at each iteration. These complexities are the best known among zeroth-order stochastic Frank-Wolfe algorithms that avoid explicit gradient calculations. Empirical experiments on convex and non-convex machine learning tasks, including sparse logistic regression, robust classification, and adversarial attacks on deep networks, validate the computational efficiency and scalability of our approach. Our algorithm demonstrates superior performance in both convergence rate and query complexity compared to existing methods.

cs.LG↗