Searcharxiv⌕ Search

arXiv subjects

Xinyi Zhang

Publications and source records attributed to Xinyi Zhang.

At least 37 records · Page 2Linked to original sources
arXiv2026

StagePilot: Stage-Level Planning for Long-Horizon Dialogue Simulation in Cybergrooming

Cybergrooming is an evolving threat to youth, requiring proactive educational interventions. We address this by modeling dialogue progression as a structured planning problem over stage-wise interactions. We propose StagePilot, a dialogue framework that separates stage-level planning from response generation, in which the model selects the next stage under constrained transitions and generates responses conditioned on it, enabling coherent and realistic progression. Reinforcement learning is used to learn stage-level policies from offline data, optimizing for both emotional alignment and goal-consistent progression. Our empirical experiments show that StagePilot generates more structured, coherent dialogue trajectories and reduces conversational stagnation compared to baselines; notably, the IQL+AWAC variant reaches the final stage more often while maintaining over 70% positive or neutral responses, yielding a 43% relative improvement.

cs.LG↗
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↗
arXiv2026

Why Database Manuals Are Not Enough: Efficient and Reliable Configuration Tuning for DBMSs via Code-Driven LLM Agents

Modern database management systems (DBMSs) expose hundreds of configuration knobs that critically influence performance. Existing automated tuning methods either adopt a data-driven paradigm, which incurs substantial overhead, or rely on manual-driven heuristics extracted from database documentation, which are often limited and overly generic. Motivated by the fact that the control logic of configuration knobs is inherently encoded in the DBMS source code, we argue that promising tuning strategies can be mined directly from the code, uncovering fine-grained insights grounded in system internals. To this end, we propose SysInsight, a code-driven database tuning system that automatically extracts fine-grained tuning knowledge from DBMS source code to accelerate and stabilize the tuning process. SysInsight combines static code analysis with LLM-based reasoning to identify knob-controlled execution paths and extract semantic tuning insights. These insights are then transformed into quantitative and verifiable tuning rules via association rule mining grounded in tuning observations. During online tuning, system diagnosis is applied to identify critical knobs, which are adjusted under the rule guidance. Evaluations demonstrate that compared to the SOTA baseline, SysInsight converges to the best configuration on average 7.11X faster while achieving a 19.9% performance improvement.

cs.DB↗
arXiv2026

Seismic full-waveform inversion based on a physics-driven generative adversarial network

Objectives: Full-waveform inversion (FWI) is a high-resolution geophysical imaging technique that reconstructs subsurface velocity models by iteratively minimizing the misfit between predicted and observed seismic data. However, under complex geological conditions, conventional FWI suffers from strong dependence on the initial model and tends to produce unstable results when the data are sparse or contaminated by noise. Methods: To address these limitations, this paper proposes a physics-driven generative adversarial network-based full-waveform inversion method. The proposed approach integrates the data-driven capability of deep neural networks with the physical constraints imposed by the seismic wave equation, and employs adversarial training through a discriminator to enhance the stability and robustness of the inversion results. Results: Experimental results on two representative benchmark geological models demonstrate that the proposed method can effectively recover complex velocity structures and achieves superior performance in terms of structural similarity (SSIM) and signal-to-noise ratio (SNR). Conclusions: This method provides a promising solution for alleviating the initial-model dependence in full-waveform inversion and shows strong potential for practical applications.

cs.LG↗
arXiv2026

MBD: A Model-Based Debiasing Framework Across User, Content, and Model Dimensions

Modern recommendation systems rank candidates by aggregating multiple behavioral signals through a value model. However, many commonly used signals are inherently affected by heterogeneous biases. For example, watch time naturally favors long-form content, loop rate favors short - form content, and comment probability favors videos over images. Such biases introduce two critical issues: (1) value model scores may be systematically misaligned with users' relative preferences - for instance, a seemingly low absolute like probability may represent exceptionally strong interest for a user who rarely engages; and (2) changes in value modeling rules can trigger abrupt and undesirable ecosystem shifts. In this work, we ask a fundamental question: can biased behavioral signals be systematically transformed into unbiased signals, under a user - defined notion of ``unbiasedness'', that are both personalized and adaptive? We propose a general, model-based debiasing (MBD) framework that addresses this challenge by augmenting it with distributional modeling. By conditioning on a flexible subset of features (partial feature set), we explicitly estimate the contextual mean and variance of the engagement distribution for arbitrary cohorts (e.g., specific video lengths or user regions) directly alongside the main prediction. This integration allows the framework to convert biased raw signals into unbiased representations, enabling the construction of higher-level, calibrated signals (such as percentiles or z - scores) suitable for the value model. Importantly, the definition of unbiasedness is flexible and controllable, allowing the system to adapt to different personalization objectives and modeling preferences. Crucially, this is implemented as a lightweight, built-in branch of the existing MTML ranking model, requiring no separate serving infrastructure.

cs.LG↗
arXiv2026

From Vulnerable to Resilient: Examining Parent and Teen Perceptions on How to Respond to Unwanted Cybergrooming Advances

Cybergrooming is a form of online abuse that threatens teens' mental health and physical safety. Yet, most prior work has focused on detecting perpetrators' behaviors, leaving a limited understanding of how teens might respond to such unwanted advances. To address this gap, we conducted an online survey with 74 participants -- 51 parents and 23 teens -- who responded to simulated cybergrooming scenarios in two ways: responses that they think would make teens more vulnerable or resilient to unwanted sexual advances. Through a mixed-methods analysis, we identified four types of vulnerable responses (encouraging escalation, accepting an advance, displaying vulnerability, and negating risk concern) and four types of protective strategies (setting boundaries, directly declining, signaling risk awareness, and leveraging avoidance techniques). As the cybergrooming risk escalated, both vulnerable responses and protective strategies showed a corresponding progression. This study contributes a teen-centered understanding of cybergrooming, a labeled dataset, and a stage-based taxonomy of perceived protective strategies, while offering implications for educational programs and sociotechnical interventions.

cs.HC↗
arXiv2026

Directed landscape convergence for the half-space log-gamma polymer $N^{2/3+δ}$ away from the boundary

We prove that the free energy of the half-space log-gamma polymer $N^{2/3+δ}$ away from the boundary in the non-attractive regime converges to the directed landscape. Based on the convergence of the full-space log-gamma free energy to the directed landscape, we couple the full-space and the half-space model and prove that the dominant contributions to free energy in both cases come from paths that remain confined to a transversal window of order $N^{2/3}$. The result follows from three main inputs: a deterministic leading-order gap between paths that deviate from the transversal window on the $N^{2/3+δ}$ scale and those within the typical $N^{2/3}$ scale; uniform exponential upper-tail bounds for half-space free energies with general slope; and existing full-space estimates on constrained and exiting free energies.

math.PR↗
arXiv2026

Stationary Log-Gamma Polymer in Half-Space

We study the half-space log-gamma polymer model with stationary initial conditions. We derive exact formulas for the distribution of the partition function along the diagonal across the entire High density phase and Low density phase. We obtain asymptotics of these distributions under the critical scaling. We also prove the first exponential upper bounds for the upper and lower tail of the scaled free energy for these half-space stationary log-gamma models.

math.PR↗
arXiv2026

Convergence of Half-Space Last Passage Percolation Away from the Boundary to the Directed Landscape

In this note, we prove convergence of the half-space exponential last passage percolation (LPP) model, away from the boundary, to the directed landscape. Our approach couples the half-space and full-space LPP models and constructs two barrier events based on the monotonicity of last passage paths. Combining this coupling with moderate deviation estimates for both models and the known convergence of full-space LPP to the directed landscape, we establish the desired convergence.

math.PR↗
arXiv2026

Integrating Large Language Models into Recommendation via Mutual Augmentation and Adaptive Aggregation

Conventional recommendation methods have achieved notable advancements by harnessing collaborative or sequential information from user behavior. Recently, large language models (LLMs) have gained prominence for their capabilities in understanding and reasoning over textual semantics, and have found utility in various domains, including recommendation. Conventional recommendation methods and LLMs each have their strengths and weaknesses. While conventional methods excel at mining collaborative information and modeling sequential behavior, they struggle with data sparsity and the long-tail problem. LLMs, on the other hand, are proficient at utilizing rich textual contexts but face challenges in mining collaborative or sequential information. Despite their individual successes, there is a significant gap in leveraging their combined potential to enhance recommendation performance. In this paper, we introduce a general and model-agnostic framework known as \textbf{L}arge \textbf{la}nguage model with \textbf{m}utual augmentation and \textbf{a}daptive aggregation for \textbf{Rec}ommendation (\textbf{Llama4Rec}). Llama4Rec synergistically combines conventional and LLM-based recommendation models. Llama4Rec proposes data augmentation and prompt augmentation strategies tailored to enhance the conventional model and LLM respectively. An adaptive aggregation module is adopted to combine the predictions of both kinds of models to refine the final recommendation results. Empirical studies on three real-world datasets validate the superiority of Llama4Rec, demonstrating its consistent outperformance of baseline methods and significant improvements in recommendation performance.

cs.IR↗
arXiv2026

PRISM: Personalized Recommendation via Information Synergy Module

Multimodal sequential recommendation (MSR) leverages diverse item modalities to improve recommendation accuracy, while achieving effective and adaptive fusion remains challenging. Existing MSR models often overlook synergistic information that emerges only through modality combinations. Moreover, they typically assume a fixed importance for different modality interactions across users. To address these limitations, we propose \textbf{P}ersonalized \textbf{R}ecommend-ation via \textbf{I}nformation \textbf{S}ynergy \textbf{M}odule (PRISM), a plug-and-play framework for sequential recommendation (SR). PRISM explicitly decomposes multimodal information into unique, redundant, and synergistic components through an Interaction Expert Layer and dynamically weights them via an Adaptive Fusion Layer guided by user preferences. This information-theoretic design enables fine-grained disentanglement and personalized fusion of multimodal signals. Extensive experiments on four datasets and three SR backbones demonstrate its effectiveness and versatility. The code is available at https://github.com/YutongLi2024/PRISM.

cs.IR↗