arXiv subjects

Chenyu Wu

Publications and source records attributed to Chenyu Wu.

At least 19 recordsLinked to original sources

arXiv2026

Evidence-Unit Fairness and the Limits of Query-Adaptive Sparse-Dense Fusion in Financial Document Retrieval

Retrieval over financial filings is difficult because queries are short and acronym-heavy while the answer-bearing evidence sits inside long, table-dense documents. We study sparse-dense hybrid retrieval on FinDER, a benchmark of expert-annotated questions over corporate 10-K filings. Our first finding is methodological: if the retrieval unit is larger than the dense encoder's input window, the dense model never sees a large share of the labeled evidence, confounding comparison against a full-text sparse baseline. We measure this directly and remove it by segmenting the corpus into encoder-sized windows. On the corrected corpus, fusing BM25 and a compact dense encoder improves reference-level Hit@10 by roughly 28 percent over either component, and training-free, untuned reciprocal rank fusion exceeds the equal-weight blend in an exploratory comparison. We then ask whether choosing the fusion weight per query helps: an oracle over the interpolation-weight grid shows headroom of 21.8 percent, yet none of the three lightweight adaptive routers (a score-confidence heuristic, a random forest over query features, and a ridge regressor over query embeddings) establishes a statistically reliable improvement over the fixed blend under company-grouped cross-validation with cluster-robust inference. Simple fusion is a strong baseline here, and we discuss why per-query weighting does not capture the available headroom.

cs.CE↗

arXiv2026

Class Weighting versus Amount Conditioning in Credit-Card Fraud Detection: A Dollar-Metric Study with a Temporal Explanation Audit

Credit-card fraud losses are monetary, but papers often judge models with transaction-level scores. We ask whether transaction amount should shape training weights or be used later to order alerts. To separate this question from ordinary class imbalance handling, we keep total fraud-case weight fixed and vary only its allocation across fraud cases. The experiments test two chronological card-fraud datasets with XGBoost under unweighted training, standard class weighting, matched log-amount weighting, stronger amount-weighted variants, and score times amount reranking. Metrics are average precision, dollar recall, and dollar precision at fixed alert budgets over five seeds, with 95 percent day-block bootstrap intervals for the main contrasts. Results are narrower than expected. Amount-derived ratio and velocity features carry much of the signal, while raw amount fields add little once those features are in the model. In the matched setting, amount-conditioned training gives only small gains over class weighting and does not consistently beat the plain unweighted model. Stronger amount weights recover more fraudulent dollars, but at lower ranking quality and dollar precision. Reranking alerts by score times amount after training gives the largest dollar-recall shift. A small SHAP audit finds larger month-to-month attribution movement for fraud cases than for aggregate traffic. In these tests, amount is useful as a feature and as an alert-ordering variable, not by itself as a better sample-weighting rule.

cs.CE↗

arXiv2026

How Early Is Early Enough? Design-Dependent Observation-Window Sufficiency in Subscription Churn Prediction

How many days of early behavior suffice for subscription churn prediction? In the public KKBox dataset, the early indicator of churn is typically an indicator of someone's contract status; however, when looking in the heavily churned manual-renewal segment, having access to early behavior creates a substantial increase in prediction for that specific segment (PR +0.10 at 120 days). A nine-window sufficiency curve shows a diminishing-returns knee in a 45-90 day band. However, stress-testing over three cohort/task designs shows that this curve is singular to the design being tested; for example, in our test with a moving target, the curve inverts and can shift depending on the feature set used. Therefore, any window-sufficiency claim should state its cohort construction, target definition, and feature families. All evidence is from one music-streaming dataset; the mechanism should generalize but the magnitudes may not.

cs.LG↗

arXiv2026

Interpretable vs Learned Encoders for High-Cardinality Fraud Detection

A total of seven categorical encoding methods were tested on the IEEE-CIS fraud benchmark dataset (590,540 records, 3.5% positives, 8 high-cardinality columns). The encoders were evaluated using a stratified 5-fold cross-validation (CV) with three repetitions. Five of the encoders had identical frozen LightGBM learners in the downstream phase, allowing for controlled comparisons of their performance to each other. CatBoost and TabNet were included as comparisons across paradigms using different learners. The entity embeddings produced the highest AUC-ROC (0.9612), with a statistically significant tie with that of CatBoost (0.9602) and statistically superior to tier group encoding (0.9548), whereas target encoding was only 0.0023 worse than tier group encoding and the auditor-friendly tier boundaries were maintained. Off-the-shelf TabNet did not outperform tree-based pipelines and collapsed under data scarcity. On AUC-PR, CatBoost leads (0.822 vs. 0.793); no encoder dominated both metrics. Per-column analysis confirmed the embedding advantage arises from joint multi-column representation.

cs.LG↗

arXiv2026

Validation-Stage Combinatorial Fusion Analysis for Imbalanced Credit-Card Fraud Detection

Credit-card fraud detection is difficult because fraudulent transactions are rare, costly, and unevenly distributed. Strong gradient-boosted tree models already perform well on structured transaction data, so the value of another fusion method is not obvious. This paper examines whether Combinatorial Fusion Analysis (CFA), which searches over model subsets and rank-score fusion rules, can still add value on the IEEE-CIS Fraud Detection benchmark. Using a leakage-free 60/20/20 train/validation/test protocol, we evaluate 480 fusion configurations built from seven base classifiers. The best test-set result comes from diversity-weighted score fusion of Random Forest, XGBoost, and LightGBM (DEF WtScore), with AUC-ROC = 0.9405, AUPRC = 0.6699, and F1 = 0.6373. Bootstrap confidence intervals from 1,000 resamples show that the gains over the strongest single model exclude zero for all three metrics. CFA matches soft voting on AUC-ROC, improves AUPRC and F1, and outperforms stacking in this setting. A CTGAN augmentation experiment gives a negative result: synthetic fraud samples degrade both individual models and CFA. Overall, CFA is most useful here not as a way to combine every classifier, but as a validation-stage method for choosing a small, complementary subset and assigning diversity-aware weights.

cs.LG↗

arXiv2026

Dynamic Interaction-Aware and Causality-Disentangled Framework for Multimodal Sentiment Analysis

Although Multimodal Sentiment Analysis (MSA) effectively leverages rich information from language, visual, and acoustic modalities, existing methods still face two core challenges: 1) static conflict suppression mechanisms fail to adapt to dynamic variations across samples, and 2) the inherent sentimental bias within the language modality, which can misguide learning from other modalities, remains entangled. To this end, we propose a Dynamic Multimodal Causal Disentanglement and Adaptive Fusion Framework (MCAF). Its cornerstone is the Multi-Granularity Causal Dynamic Router and a Conditional Diffusion Denoising Module. First, we introduce a causal intervention module based on the information bottleneck principle, which builds a Structural Causal Model to disentangle sentimental bias from language features, yielding a "de-confounded" language representation as a pure guiding signal. Second, we devise a Dynamic Multimodal Router that evaluates the interaction states (complementary, conflicting, or redundant) among visual, acoustic, and de-confounded language signals in real-time across three levels: feature, temporal, and modality, then adaptively allocates weights and routes information flow for fine-grained regulation. Finally, a lightweight Conditional Diffusion Denoising Module performs iterative denoising on the fused joint representation to explicitly filter out residual irrelevant information, generating a robust hyper-modality representation. Extensive experiments on the CMU-MOSI and CMU-MOSEI benchmarks show that MCAF sets new state-of-the-art on key classification metrics, achieving an Acc-2/F1 of 86.52%/86.51% on MOSI and 86.72%/86.65% on MOSEI, while remaining highly competitive on others. Comprehensive analyses and visualizations further validate its efficacy in dynamically perceiving interactions, disentangling bias, and enhancing interpretability.

cs.MM↗

arXiv2026

RiGS: Rigid-aware 4D Gaussian Splatting from a Single Monocular Video

Reconstructing dynamic 3D scenes from monocular videos is a fundamental yet highly challenging task, as real-world motions often involve both long-term smooth transformations and short-term complex deformations. Existing methods either struggle to maintain temporal consistency or fail to capture high-frequency dynamics due to limited motion modeling capacity. In this work, we present Rigid-aware 4D Gaussian Splatting (RiGS), which simultaneously captures motions across multiple temporal scales. Specifically, RiGS introduces three types of Gaussian primitives: static, rigid, and transient, which represent static backgrounds, long-term low-frequency motions, and short-term high-frequency dynamics, respectively. An object-wise dynamic mask is proposed to aggregate long-range spatiotemporal motion information and guide the decomposition of static and dynamic regions. To jointly model motion across scales, rigid Gaussians are allowed to transition into transient Gaussians based on their temporal duration, and both are optimized under scene flow guidance, providing dense 3D motion supervision. Extensive experiments demonstrate that RiGS achieves state-of-the-art performance on novel view synthesis benchmarks. Code is available at \hyperlink{https://github.com/ladvu/RiGS}{https://github.com/ladvu/RiGS}.

cs.CV↗

arXiv2026

Skill-Conditioned Visual Geolocation for Vision-Language Models

Vision-language models (VLMs) have shown a promising ability in image geolocation, but they still lack structured geographic reasoning and the capacity for autonomous self-evolution. Existing methods predominantly rely on implicit parametric memory, which often exploits outdated knowledge and generates hallucinated reasoning. Furthermore, current inference is a "one-off" process, lacking the feedback loops necessary for self-evolution based on reasoning outcomes. To address these issues, we propose GeoSkill, a training-free framework based on an evolving Skill-Graph. We first initialize the graph by refining human expert trajectories into atomic, natural-language skills. For execution, GeoSkill employs an inference model to perform direct reasoning guided by the current Skill-Graph. For continuous growth, an Autonomous Evolution mechanism leverages a larger model to conduct multiple reasoning rollouts on image-coordinate pairs sourced from web-scale data and verified real-world reasoning. By analyzing both successful and failed trajectories from these rollouts, the mechanism iteratively synthesizes and prunes skills, effectively expanding the Skill-Graph and correcting geographic biases without any parameter updates. Experiments demonstrate that GeoSkill achieves promising performance in both geolocation accuracy and reasoning faithfulness on GeoRC, while maintaining superior generalization across diverse external datasets. Furthermore, our autonomous evolution fosters the emergence of novel, verifiable skills, significantly enhancing the system's cognition of real-world geographic knowledge beyond isolated case studies.

cs.CV↗

arXiv2026

ActivityEditor: Learning to Synthesize Physically Valid Human Mobility

Human mobility modeling is indispensable for diverse urban applications. However, existing data-driven methods often suffer from data scarcity, limiting their applicability in regions where historical trajectories are unavailable or restricted. To bridge this gap, we propose \textbf{ActivityEditor}, a novel dual-LLM-agent framework designed for zero-shot cross-regional trajectory generation. Our framework decomposes the complex synthesis task into two collaborative stages. Specifically, an intention-based agent, which leverages demographic-driven priors to generate structured human intentions and coarse activity chains to ensure high-level socio-semantic coherence. These outputs are then refined by editor agent to obtain mobility trajectories through iteratively revisions that enforces human mobility law. This capability is acquired through reinforcement learning with multiple rewards grounded in real-world physical constraints, allowing the agent to internalize mobility regularities and ensure high-fidelity trajectory generation. Extensive experiments demonstrate that \textbf{ActivityEditor} achieves superior zero-shot performance when transferred across diverse urban contexts. It maintains high statistical fidelity and physical validity, providing a robust and highly generalizable solution for mobility simulation in data-scarce scenarios. Our code is available at: https://anonymous.4open.science/r/ActivityEditor-066B.

cs.AI↗

arXiv2026

Exploiting Self-Sustainable Information-Bearing RIS in Underlay CR-NOMA Networks

Information-bearing reconfigurable intelligent surfaces (IB-RIS) provide a promising solution to self-sustainable and green communications by harvesting ambient radio frequency energy while embedding information via passive reflection. This paper investigates a self-sustainable IB-RIS (SIB-RIS)-assisted non-orthogonal multiple access (NOMA) network operating in an underlay cognitive radio (CR) system. Specifically, a multi-antenna primary transmitter (PT) serves a primary user (PU) and concurrently illuminates the secondary nodes, which enables each SIB-RIS to perform simultaneous energy harvesting and backscatter-based information embedding at each RIS. Based on this model, a weighted sum spectral efficiency (WSSE) maximization problem is formulated for the secondary network by jointly optimizing the PT transmit beamforming vector, the SIB-RIS reflection coefficients, and the power-splitting ratios. To tackle the intricately-coupled non-convex problem, an efficient block coordinate descent (BCD) optimization framework is developed, which leverages fractional programming via Lagrangian dual and quadratic transforms together with a difference-of-convex programming approach. Numerical results demonstrate that the proposed SIB-RIS-assisted NOMA CR system yields substantial WSSE gains over both orthogonal multiple access (OMA)-based and active antenna schemes. Moreover, a 2-bit discrete-phase SIB-RIS implementation achieves competitive to which WSSE performance, confirming the practicality of the low-resolution architecture.

eess.SP↗

arXiv2025

A Cartesian-3j Framework for Machine Learning Interatomic Potentials

Machine learning interatomic potentials (MLIPs) have brought substantial gains in the extrapolation capability in computational chemistry. However, most equivariant models are typically built with spherical tensors (STs), while Cartesian tensor formulations remain less developed despite their natural alignment with atomic coordinates and tensorial targets. In this work, we develop a Cartesian framework for irreducible Cartesian tensors (ICTs) by introduce the \texttt{Cartesian-3j} symbol and Cartesian Generalized Clebsch-Gordan Coefficients, which serve as direct analogues of the \texttt{Wigner-3j} symbol and Generalized Clebsch-Gordan coefficients defined for ST coupling. We extend the \texttt{e3nn} library to support ICT product, and use this framework to build Cartesian counterparts of \texttt{MACE}, \texttt{NequIP}, and \texttt{Allegro}, allowing the first controlled comparison where architectures are held fixed and only the tensor basis is changed. Our experiments show that irreducible Cartesian models can achieve accuracy comparable to spherical counterparts, but direct Cartesianization incurs unfavorable compute and memory scaling, motivating dedicated Cartesian architectural choices. Leveraging ICTs and our framework, we introduce \texttt{TACE-v1-OAM-M} and demonstrate that it achieves competitive performance on Matbench Discovery compared to state-of-the-art ST models.

physics.chem-ph↗

arXiv2025

Mini-SFC: A Comprehensive Simulation Framework for Orchestration and Management of Service Function Chains

In the continuously evolving cloud computing and network environment, service function chain (SFC) plays a crucial role in implementing complex services in the network with its flexible deployment capabilities. To address the limitations of existing SFC simulation tools, this paper introduces Mini-SFC, a modular simulation framework that supports both numerical and container-based virtual simulations, while also supporting online dynamic topology adjustments. As an open-source platform emphasizing user-friendliness, Mini-SFC facilitates rapid algorithm verification and realistic service deployment validation. By simplifying module design and providing standardized solver interfaces, Mini-SFC significantly shortens the learning curve for researchers and enhances the flexibility and scalability required for advanced SFC management and optimization. For readers interested in exploring or utilizing Mini-SFC, more information is available on the official project page.

cs.SE↗

arXiv2025

Data-driven Augmentation of a Turbulence Model in Three dimensional Separated Flows

Classic turbulence models often struggle to accurately predict complex flows. Although data-driven techniques have addressed these shortcomings, most existing research has concentrated on two-dimensional (2D) cases. This study bridges this gap by enhancing a data-driven turbulence model, the SST-CND (shear stress transport-conditioned) model, which was originally trained on 2D separated flows, in 3D scenarios. An additional correction term, \b{eta}_3D, is introduced to account for 3D effects. The distribution of this term is determined through a 3D field inversion process using high-fidelity data obtained from the flow around a cube. An algebraic expression for \b{eta}_3D is then derived through symbolic regression and formulated to degrade to zero in 2D cases. The performance of the resulting SST-CND3D model is evaluated across a range of flows. In 2D flows, the SST-CND3D model performs identically to its 2D-trained predecessor. However, the model exhibits superior performance in 3D flows, such as the flow around the complex JAXA standard model high-lift configuration. These findings indicate that a sequential approach, constructing a 3D correction term that vanishes in 2D on top of a 2D-trained model, constitutes a promising method for developing data-driven turbulence models that perform accurately in 3D while preserving their effectiveness in 2D.

physics.flu-dyn↗

arXiv2025

BERTector: An Intrusion Detection Framework Constructed via Joint-dataset Learning Based on Language Model

Intrusion detection systems (IDS) are widely used to maintain the stability of network environments, but still face restrictions in generalizability due to the heterogeneity of network traffics. In this work, we propose BERTector, a new framework of joint-dataset learning for IDS based on BERT. BERTector integrates three key components: NSS-Tokenizer for traffic-aware semantic tokenization, supervised fine-tuning with a hybrid dataset, and low-rank adaptation for efficient fine-tuning. Experiments show that BERTector achieves state-of-the-art detection accuracy, strong generalizability, and excellent robustness. BERTector achieves the highest accuracy of 99.28% on NSL-KDD and reaches the average 80% detection success rate against four perturbations. These results establish a unified and efficient solution for modern IDS in complex and dynamic network environments.

cs.CR↗

arXiv2025

GeoMAE: Masking Representation Learning for Spatio-Temporal Graph Forecasting with Missing Values

The ubiquity of missing data in urban intelligence systems, attributable to adverse environmental conditions and equipment failures, poses a significant challenge to the efficacy of downstream applications, notably in the realms of traffic forecasting and energy consumption prediction. Therefore, it is imperative to develop a robust spatio-temporal learning methodology capable of extracting meaningful insights from incomplete datasets. Despite the existence of methodologies for spatio-temporal graph forecasting in the presence of missing values, unresolved issues persist. Primarily, the majority of extant research is predicated on time-series analysis, thereby neglecting the dynamic spatial correlations inherent in sensor networks. Additionally, the complexity of missing data patterns compounds the intricacy of the problem. Furthermore, the variability in maintenance conditions results in a significant fluctuation in the ratio and pattern of missing values, thereby challenging the generalizability of predictive models. In response to these challenges, this study introduces GeoMAE, a self-supervised spatio-temporal representation learning model. The model is comprised of three principal components: an input preprocessing module, an attention-based spatio-temporal forecasting network (STAFN), and an auxiliary learning task, which draws inspiration from Masking AutoEncoders to enhance the robustness of spatio-temporal representation learning. Empirical evaluations on real-world datasets demonstrate that GeoMAE significantly outperforms existing benchmarks, achieving up to 13.20\% relative improvement over the best baseline models.

cs.LG↗

arXiv2025

Interplay Between AI and Space-Air-Ground Integrated Network: The Road Ahead

Space-air-ground integrated network (SAGIN) is envisioned as a key network architecture for achieving ubiquitous coverage in the next-generation communication system. Concurrently, artificial intelligence (AI) plays a pivotal role in managing the complex control of SAGIN, thereby enhancing its automation and flexibility. Despite this, there remains a significant research gap concerning the interaction between AI and SAGIN. In this context, we first present a promising approach for developing a generalized AI model capable of executing multiple tasks simultaneously in SAGIN. Subsequently, we propose a framework that leverages software-defined networking (SDN) and AI technologies to manage the resources and services across the entire SAGIN. Particularly, we demonstrate the real-world applicability of our proposed framework through a comprehensive case study. These works pave the way for the deep integration of SAGIN and AI in future wireless networks.

cs.NI↗

arXiv2025

Super-resolution Wideband Beam Training for Near-field Communications with Ultra-low Overhead

In this paper, we propose a super-resolution wideband beam training method for near-field communications, which is able to achieve ultra-low overhead. To this end, we first study the multi-beam characteristic of a sparse uniform linear array (S-ULA) in the wideband. Interestingly, we show that this leads to a new beam pattern property, called rainbow blocks, where the S-ULA generates multiple grating lobes and each grating lobe is further splitted into multiple versions in the wideband due to the well-known beam-split effect. As such, one directional beamformer based on S-ULA is capable of generating multiple rainbow blocks in the wideband, hence significantly extending the beam coverage. Then, by exploiting the beam-split effect in both the frequency and spatial domains, we propose a new three-stage wideband beam training method for extremely large-scale array (XL-array) systems. Specifically, we first sparsely activate a set of antennas at the central of the XL-array and judiciously design the time-delay (TD) parameters to estimate candidate user angles by comparing the received signal powers at the user over subcarriers. Next, to resolve the angular ambiguity introduced by the S-ULA, we activate all antennas in the central subarray and design an efficient subcarrier selection scheme to estimate the true user angle. In the third stage, we resolve the user range at the estimated user angle with high resolution by controlling the splitted beams over subcarriers to simultaneously cover the range domain. Finally, numerical results are provided to demonstrate the effectiveness of proposed wideband beam training scheme, which only needs three pilots in near-field beam training, while achieving near-optimal rate performance.

eess.SP↗

arXiv2025

Numerical Simulation of Three-dimensional High-Lift Configurations Using Data-Driven Turbulence Model

Traditional Reynolds-averaged Navier-Stokes (RANS) equations often struggle to predict separated flows accurately. Recent studies have employed data-driven methods to enhance predictions by modifying baseline equations, such as field inversion and machine learning (FIML) with symbolic regression. However, data-driven turbulence models exhibit limited adaptability and are rarely applied to complex engineering problems. This study examines the application of data-driven turbulence models to complex three-dimensional high-lift configurations, extending their usability beyond previous applications. First, the generalizability of the SST-CND model, derived from conditioned field inversion and symbolic regression, is validated. Then, the spatially varying correction factor obtained through conditioned field inversion is transferred to the three-equation k-(v^2)-w model. The 30P30N three-element airfoil, the JAXA Standard Model (JSM), and the high-lift version of the NASA Common Research Model (CRM-HL) are numerically simulated. The results indicated that the SST-CND model significantly improves the prediction of stall characteristics, demonstrating satisfactory generalizability. The corrected k-(v^2 )-w-CND model accurately predicts the stall characteristics of CRM-HL, with a relative error of less than 5% compared to experimental results. This confirms the strong transferability of the model correction derived from conditioned field inversion across different turbulence models.

physics.flu-dyn↗