Searcharxiv⌕ Search

arXiv subjects

Xinyi Zhang

Publications and source records attributed to Xinyi Zhang.

At least 19 recordsLinked to original sources
arXiv2026

SG-UMP: Sequence-Guided Universal Multimodal Prioritization Calculation Framework

Multimodal sequential recommendation (MSR) improves recommendation by incorporating heterogeneous information such as text, images, and user interactions. However, existing MSR methods often fail to capture user-level preference heterogeneity and dataset-level modality bias, limiting their adaptability across users and datasets. To address this issue, we propose \textbf{S}equence-\textbf{G}uided \textbf{U}niversal \textbf{M}ultimodal \textbf{P}rioritization Calculation Framework (\textbf{SG-UMP}), a plug-and-play plugin for enhancing multimodal information processing in MSR. SG-UMP includes a Module Combiner for flexible multimodal processing and a Module Router for dynamic module ordering, enabling adaptation to both user preferences and dataset characteristics. Experiments on four real-world datasets show that SG-UMP consistently improves recommendation performance across different backbones and multimodal settings. The code is available at https://github.com/esemsc-xz524/SG-UMP .

cs.IR↗
arXiv2026

StartupBench: Benchmarking General-Purpose Agents on Market-Validated End-to-End Workflows

Recent advances in Large Language Models(LLMs) and agents have substantially improved the ability of AI systems to execute complex tasks. Yet existing benchmarks largely rely on researcher-selected tasks, leaving uncertain whether such progress extends to the work that real-world users actually demand from AI systems. We introduce \textbf{StartupBench}, an E2E agent benchmark grounded in market-validated AI startup products. Rather than defining tasks from pre-defined assumptions about useful agent capabilities, we systematically study AI products with demonstrated adoption, together with their product workflows and users, to identify real-world tasks for which AI has established practical demand across diverse professional domains. We translate these workflows into complete deliverable-oriented tasks and evaluate them with fine-grained rubrics capturing their complex requirements. Across representative models evaluated under a unified agent harness, even the strongest model successfully completes only approximately 30\% of StartupBench, despite making substantial partial progress on many tasks. Further analysis identifies aspects like complex instruction following and domain-specific expertise as major sources of failure. Our results reveal that many market-validated workflows remain beyond the reliable capabilities of current general-purpose agents, establishing StartupBench as an empirical measure of progress toward E2E completions of real-world user tasks.

cs.AI↗
arXiv2026

GAINS: Leveraging Inconsistent Human Intervention Signals in Reinforcement Learning

Correcting robot manipulation policies through human intervention holds great promise for real-world deployment, yet human operators are inherently imperfect in both the actions they provide and the timing of their intervention signals. While the former has been extensively discussed in reinforcement learning (RL), the latter remains underexplored. At high control frequencies, human intervention signals are often delayed and inconsistent across time and state space. In this work, we present GAINS, a framework for leveraging inconsistent human intervention signals in RL. At the core of GAINS, we employ distributional RL with quantile Q-networks to model the return variability induced by sparse task rewards and inconsistent human interventions. Building on this distributional representation, we introduce a pessimistic exploration strategy that promotes safe and sample-efficient learning under human corrections. We evaluate GAINS on four diverse simulated manipulation tasks and two challenging real-world scenarios against state-of-the-art intervention-based methods. GAINS achieves a 22% higher task success rate than RLIF and improves recovery success by up to 43% in failure scenarios. These results highlight the importance of modeling return variability induced by human imperfection for real-world deployment of intervention-based learning.

cs.RO↗
arXiv2026

Bias Mitigation in Face Recognition via Demographic-based Supervised Contrastive Learning

Face recognition systems have been shown to be biased toward certain demographic groups by exhibiting different error rates across gender, age, or ethnicity. Though the imbalance of the training data with respect to these demographics is one cause of this bias, training on artificially balanced groups does not completely mitigate the problem. For deployment, face recognition typically works at operating points allowing very low false match rates and, hence, on the tail of the non-match score distribution. While class balancing can improve the means of these distributions, the aim of our approach is to improve fairness by addressing the behavior in the tail. Particularly, we propose the Demographic-based Supervised Contrastive loss (DeSCon) for face recognition, which relies on a well-designed composition of training batches and demographic-aware pair selection. Our experimental evaluation on both demographically-labeled datasets and standard verification benchmarks shows that DeSCon can improve fairness beyond balancing training datasets while maintaining competitive verification performance. Source code is available upon request.

cs.CV↗
arXiv2026

ASIDE: From Conflict Participants to Co-Observers Through Dyadic Spectator Reflection

When two people argue over text, each knows what they meant and can only guess what the other was thinking. Existing AI reflection tools work from one person's account, and dyadic tools support co-expression without making the gap between accounts inspectable. We propose Dyadic Spectator Reflection (DSR), an interaction structure in which both partners externalize their models of each other independently and then encounter them together, and present ASIDE, a system that operationalizes it. ASIDE replays a past text conflict as a pixel-art theatrical scene where each character's unspoken state appears as an AI-inferred thought bubble either partner can contest and rewrite. Each edits alone, and the two versions meet only when both are done, in a scene they watch together. In an exploratory study, 10 couples revisited real conflicts and described the scene as a shared position from which to observe their own argument, stepping out of their roles without disengaging from it, and Divergence Cards as a way to locate specific interpretation gaps afterward. We contribute DSR as a reusable interaction structure, ASIDE as its system realization, and exploratory empirical findings on how couples used it.

cs.HC↗
arXiv2026

Noise-Aware Shrinkage for Differentially Private Zeroth-Order Fine-Tuning of Large Language Models

Differentially private zeroth-order optimization (DP-ZO) enables memory-efficient private fine-tuning of large language models using only forward evaluations. Existing aggregation-based DP-ZO methods reconstruct model updates at a fixed scale, ignoring that the strength of useful signals varies throughout training. Consequently, noise-dominated updates may receive excessive weight and degrade model utility. To address this issue, we propose SAGE, a noise-aware shrinkage method that adaptively attenuates privatized estimates according to their estimated signal quality. SAGE subtracts the known Gaussian noise variance from the observed second moment to estimate the underlying signal energy, stabilizes this estimate through temporal tracking, and compares its current signal-to-noise level with a warm-up reference to derive a bounded shrinkage factor. As pure post-processing, SAGE requires neither additional privacy budget nor model queries and introduces only constant additional state. Our theoretical analysis shows that shrinkage reduces the quadratic update-risk term faster than the linear descent term, preserving useful descent while limiting the influence of noise-dominated updates. Experiments on RoBERTa-large, OPT-1.3B, and OPT-6.7B demonstrate that SAGE outperforms existing baselines in most settings under the same privacy budgets while preserving the forward-only memory efficiency of DP-ZO.

cs.LG↗
arXiv2026

TactiPlay: Multi-Granularity Tactical Parsing and Video-Anchored Match Review for Amateur Badminton Players

Amateur badminton players increasingly record matches, yet existing tools provide only aggregate statistics or generic summaries, leaving most unable to extract tactical insights without expert guidance. A formative study (N=8) reveals the need for multi-granularity, video-anchored tactical analysis centered on rallies. We derive a taxonomy of performance issues from national-level athletes' annotations and present TactiPlay, an interactive system that instantiates an expert-taxonomy-guided, rally-level, video-anchored review workflow. The system's analytical pipeline organizes match events into taxonomy-grounded feedback, while its interface links structured reports to rally summaries, video evidence, and court visualizations. A within-subjects study (N=16) shows that TactiPlay elicits more frequent, concrete, actionable, and appropriate reflections than a report-and-statistics baseline. These findings show how organizing reviewed match evidence into a taxonomy-guided, video-linked workflow can support amateur players' tactical reflection.

cs.HC↗
arXiv2026

Antichiral hinge states in a higher-order photonic nodal ring semimetal

Antichiral states propagate in the same direction on opposite boundaries, defying the conventional constraint that boundary modes must cancel net chirality. Previously found antichiral states have been limited to first-order topological semimetals. However, antichiral hinge states, the antichiral counterpart of recently discovered higher-order chiral hinge states, remain elusive. Here, we report the observation of antichiral hinge states in a higher-order nodal-ring semimetal made of a three-dimensional gyromagnetic photonic crystal. Near-field scanning measurements reveal a pair of hinge states at two parallel one-dimensional boundaries propagating unidirectionally along the same direction, with robust transport against metallic scatterers. Their spatial positions can be reconfigured by adding or removing photonic layers. The additionally observed antichiral and drumhead surface states manifest a hierarchy of first- and second-order topological boundary states within a single photonic system. Our work extends antichiral states to higher-order topological semimetals and has potential applications in robust and reconfigurable photonic routing.

physics.optics↗
arXiv2026

Hitting time mixing for random $k$-cycles

In this paper, we study the random walk on the symmetric group $\mathfrak{S}_n$ generated by the conjugacy class of $k$-cycles, where $2\le k=o(n/(\log n)^4)$. We prove that the walk exhibits hitting-time mixing: at the first time when every card has been touched, the distribution is already close to equilibrium. For odd $k$, the equilibrium measure is the uniform measure on $\mathfrak{A}_n$. For even $k$, the walk first mixes to the parity mixture determined by the hitting time, and in our range this mixture is asymptotically $U_{\mathfrak{S}_n}$. Our argument combines a refined fixed-time approximation for the random $k$-cycle walk near the cutoff window with an auxiliary marking scheme inspired by Jain-Sawhney's work (arXiv:2410.23944) on random transpositions. The main new feature is a parity-compatible coupling which handles both odd and even $k$-cycles in a unified framework. We also prove a hitting-time mixing result in the opposite regime $k\ge n-o(n^{1/2})$, and formulate a conjecture for all $2\le k\le n-1$.

math.PR↗
arXiv2026

From Blind Search to Memory-Aware Evolution: Efficient DBMS Tuning via Collaborative Diagnosis and Utility-Aware Retrieval

Modern DBMSs expose multiple configurable components (e.g., knobs, query hints, and indexes) that jointly determine query performance. Multi-component tuning is challenging due to the large combinatorial search space and the difficulty of learning effective tuning policies under limited feedback. Existing approaches still rely on blind search over the configuration space and interaction-heavy policy learning, leading to high tuning overhead and limited performance gains. Recent advances in large language models (LLMs) enable knowledge-driven tuning, but existing LLM-based methods fail to effectively exploit online feedback and historical observations, often converging prematurely to suboptimal configurations. In this paper, we present EvoTune, a memory-aware evolution framework for multi-component DBMS tuning. EvoTune first localizes a query-specific high-impact subspace via collaborative diagnosis, which combines lightweight pattern learning with LLM-based reasoning. It further introduces a utility-aware retrieval policy that selects informative observations based on their resulting long-term performance improvement, instead of similarity-based retrieval. To support continual improvement, EvoTune organizes tuning feedback into a hierarchical memory and incrementally refines both subspace localization and tuning policies without requiring LLM fine-tuning. Extensive experiments show that EvoTune consistently outperforms state-of-the-art baselines, achieving up to 44.5% performance improvement under the same tuning budget and reaching the best competing baseline's final performance up to 3.9X faster.

cs.DB↗
arXiv2026

Teach Multimodal Recommendation Model to See via Personalized Visual Extraction and Adaptive Learning

Multimodal sequential recommendation (MSR) incorporates textual and visual information to improve recommendation quality. However, recent studies and our empirical analysis show that visual features are often underutilized, thereby contributing far less than textual signals. We attribute this issue to two factors: insufficient visual representation learning (pretrained encoders fail to capture preference-relevant cues) and unbalanced visual-text optimization (textual features dominate the learning process). To address these issues, we propose Teach Multimodal Recommendation Model to See via Personalized Visual Extraction and Adaptive Learning (REVEAL), a plug-and-play framework that enhances visual representation learning and cross-modal optimization without modifying the original recommendation backbone. REVEAL consists of Feedback-Guided Visual Extraction (FVE), which refines prompt-guided visual extraction through task-level feedback, and Adaptive Visual Learning (AVL), which dynamically reweights visual learning to alleviate modality imbalance. Experiments on multiple real-world datasets and MSR backbones demonstrate that REVEAL consistently improves recommendation performance. Further analysis shows that these gains arise from more effective attention to preference-relevant visual regions and better visual utilization during training. The code is available at https://github.com/YutongLi2024/REVEAL.

cs.IR↗
arXiv2026

Fine-Tuned LLM as a Complementary Predictor Improving Ads System

Recommendation systems power engagement and monetization across feeds, ads, and short-video platforms, but translating the latest advances in Large Language Models into Recommendation Systems (RecSys) gains remains rare, particularly in advertising and production-scale real-world industry setups. Prior real-world LLM successes typically fall into three buckets: (a) generative retrieval that directly predicts the next items for candidate generation, (b) late-stage re-ranking that uses LLMs, and (c) auxiliary signal enrichment with LLMs. We introduce a complementary paradigm for ads: a fine-tuned open-source LLM used not as a ranker, but as an ads-specific ancillary predictor, forecasting likely advertisers from user profiles and histories. This LLM-driven advertiser prediction augments conventional candidate generation and provides informative priors to downstream ranking. Developed in a large-scale production advertising system, our approach produces substantial offline improvements and measurable online business impact, demonstrating that LLM world knowledge and predictive capacity can be efficiently harnessed. Beyond validating LLMs for ads applications, our results show that targeted ancillary predictions can unlock end-to-end gains across both retrieval and late-stage ranking, offering a practical path to LLM-enhanced recommendation at scale.

cs.IR↗
arXiv2026

A Workflow for Evaluating Regional Treatment Effect Heterogeneity in Multi-Regional Clinical Trials

Multi-regional clinical trials (MRCTs) enable efficient global drug development by assessing treatment effects across regions within a single protocol. While powered for overall efficacy, MRCTs are typically not designed to provide confirmatory evidence on regional differences, making an assessment of observed regional heterogeneity largely exploratory and susceptible to sampling variability. Despite this challenge, understanding regional heterogeneity remains important for interpretation and regulatory decision-making. This paper proposes a structured, question-driven framework to guide exploratory assessments of regional heterogeneity in MRCTs. We formulate four key questions to clarify the objectives of such analyses and propose a set of statistical methods to address them. Simulation studies evaluate performance under scenarios with no heterogeneity and heterogeneity driven by observed or unobserved treatment effect modifiers, illustrating how a structured approach can support transparent and cautious interpretation.

stat.AP↗
arXiv2026

FAME: Feature Activation Map Explanation on Image Classification and Face Recognition

Deep Learning has revolutionized machine learning, reaching unprecedented levels of accuracy, but at the cost of reduced interpretability. Especially in image processing systems, deep networks transform local pixel information into more global concepts in a highly obscured manner. Explainable AI methods for image processing try to shed light on this issue by highlighting the regions of the image that are important for the prediction task. Among these, Class Activation Mapping (CAM) and its gradient-based variants compute attributions based on the feature map and upscale them to the image resolution, assuming that feature map locations are influenced only by underlying regions. Perturbation-based methods, such as CorrRISE, on the other hand, try to provide pixel-level attributions by perturbing the input with fixed patches and checking how the output of the network changes. In this work, we propose Feature Activation Map Explanation (FAME), which combines both worlds by using network gradients to compute changes to the input image, manipulating it in a gradient-driven way rather than using fixed patches. We apply this technique on two common tasks, image classification and face recognition, and show that CAM's above-mentioned assumption does not hold for deeper networks. We qualitatively and quantitively show that FAME produces attribution maps that are competitive state-of-the-art systems. Our code is available: {\footnotesize https://github.com/AIML-IfI/fame.}

cs.CV↗
arXiv2026

RL-RIG: A Generative Spatial Reasoner via Intrinsic Reflection

Recent advancements in image generation have achieved impressive results in producing high-quality images. However, existing image generation models still generally struggle with a spatial reasoning dilemma, lacking the ability to accurately capture fine-grained spatial relationships from the prompt and correctly generate scenes with structural integrity. To mitigate this dilemma, we propose RL-RIG, a Reinforcement Learning framework for Reflection-based Image Generation. Our architecture comprises four primary components: Diffuser, Checker, Actor, and Inverse Diffuser, following a Generate-Reflect-Edit paradigm to spark the Chain of Thought reasoning ability in image generation for addressing the dilemma. To equip the model with better intuition over generation trajectories, we further develop Reflection-GRPO to train the VLM Actor for edit prompts and the Image Editor for better image quality under a given prompt, respectively. Unlike traditional approaches that solely produce visually stunning yet structurally unreasonable content, our evaluation metrics prioritize spatial accuracy, utilizing Scene Graph IoU and employing a VLM-as-a-Judge strategy to assess the spatial consistency of generated images on LAION-SG dataset. Experimental results show that RL-RIG outperforms existing state-of-the-art open-source models by up to 11% in terms of controllable and precise spatial reasoning in image generation.

cs.CV↗
arXiv2026

HIRE: A Hybrid Learned Index for Robust and Efficient Performance under Mixed Workloads

Indexes are critical for efficient data retrieval and updates in modern databases. Recent advances in machine learning have led to the development of learned indexes, which model the cumulative distribution function of data to predict search positions and accelerate query processing. While learned indexes substantially outperform traditional structures for point lookups, they often suffer from high tail latency, suboptimal range query performance, and inconsistent effectiveness across diverse workloads. To address these challenges, this paper proposes HIRE, a hybrid in-memory index structure designed to deliver efficient performance consistently. HIRE combines the structural and performance robustness of traditional indexes with the predictive power of model-based prediction to reduce search overhead while maintaining worst-case stability. Specifically, it employs (1) hybrid leaf nodes adaptive to varying data distributions and workloads, (2) model-accelerated internal nodes augmented by log-based updates for efficient updates, (3) a nonblocking, cost-driven recalibration mechanism for dynamic data, and (4) an inter-level optimized bulk-loading algorithm accounting for leaf and internal-node errors. Experimental results on multiple real-world datasets demonstrate that HIRE outperforms both state-of-the-art learned indexes and traditional structures in range-query throughput, tail latency, and overall stability. Compared to state-of-the-art learned indexes and traditional indexes, HIRE achieves up to 41.7$\times$ higher throughput under mixed workloads, reduces tail latency by up to 98% across varying scenarios.

cs.DB↗
arXiv2026

Xpertbench: Expert Level Tasks with Rubrics-Based Evaluation

As Large Language Models (LLMs) exhibit plateauing performance on conventional benchmarks, a pivotal challenge persists: evaluating their proficiency in complex, open-ended tasks characterizing genuine expert-level cognition. Existing frameworks suffer from narrow domain coverage, reliance on generalist tasks, or self-evaluation biases. To bridge this gap, we present XpertBench, a high-fidelity benchmark engineered to assess LLMs across authentic professional domains. XpertBench consists of 1,346 meticulously curated tasks across 80 categories, spanning finance, healthcare, legal services, education, and dual-track research (STEM and Humanities). These tasks are derived from over 1,000 submissions by domain experts--including researchers from elite institutions and practitioners with extensive clinical or industrial experience--ensuring superior ecological validity. Each task uses detailed rubrics with mostly 15-40 weighted checkpoints to assess professional rigor. To facilitate scalable yet human-aligned assessment, we introduce ShotJudge, a novel evaluation paradigm that employs LLM judges calibrated with expert few-shot exemplars to mitigate self-rewarding biases. Our empirical evaluation of state-of-the-art LLMs reveals a pronounced performance ceiling: even leading models achieve a peak success rate of only ~66%, with a mean score around 55%. Models also exhibit domain-specific divergence, showing non-overlapping strengths in quantitative reasoning versus linguistic synthesis.. These findings underscore a significant "expert-gap" in current AI systems and establish XpertBench as a critical instrument for navigating the transition from general-purpose assistants to specialized professional collaborators.

cs.AI↗
arXiv2026

Probing Red Giant Interiors with G-Dominated Mixed Modes I: The Cases of KIC 9145955, KIC 9970396, KIC 9882316 and KIC 11968334

We perform a detailed asteroseismic analysis of four red giants observed by Kepler: KIC 9145955, KIC 9970396, KIC 9882316, and KIC 11968334. Our study is based on individual oscillation frequencies, with particular emphasis on gravity-dominated (g-dominated) mixed modes. These modes are highly sensitive to the deep stellar interior and serve as powerful diagnostics of core structure, convective overshooting, and internal rotation. Moreover, surface effects have minimal impact on g-dominated mixed modes. To ensure accurate frequency matching between observations and theoretical models, we apply a mode-identification technique that effectively distinguishes p-dominated from g-dominated modes. Although a definitive confirmation of this trend requires a substantially larger asteroseismic sample, our best-fitting models suggest that the derived convective overshooting parameter ($f_{ov}$) increases with stellar mass. In particular, within our sample the star with a mass exceeding $1.4M_{\odot}$ requires $f_{ov} > 0.01$, whereas lower-mass red giants tend to have $f_{ov}$ <0.01. In addition, the average core rotation rate of KIC 11968334 is precisely determined to be $0.7409\pm0.0113 μ$Hz from the asteroseismic model.

astro-ph.SR↗