arXiv subjects

Chong Wu

Publications and source records attributed to Chong Wu.

At least 19 recordsLinked to original sources

arXiv2026

Intrinsic Low-Tucker-Rank Theory and Unified Tensor CUR Decomposition for High-Dimensional Hyperinterpolation

High-dimensional hyperinterpolation is severely hampered by the curse of dimensionality, as its coefficient tensors grow exponentially with the ambient dimension. Existing research predominantly focuses on heuristic algorithmic optimizations, often overlooking the inherent structural properties of these tensors. This paper establishes a rigorous theory of intrinsic low-$\epsilon$-Tucker-rank for hyperinterpolation coefficient tensors, delivering near-optimal low-rank approximations with error bounds that are nearly independent of the dimension. We further construct a unified, Tucker-compatible theoretical framework that integrates both Chidori-type and Fiber-type tensor CUR (TCUR) decompositions, deriving tight and stable Frobenius-norm error estimates that depend exclusively on tensor spectral properties and index set geometry. We mathematically verify the convergence and numerical stability of greedy adaptive index selection schemes and prove their near-optimality, enabling a fully tensor-free hyperinterpolation workflow that avoids constructing the full coefficient array. Three practical greedy TCUR algorithms and a lightweight TCUR-to-Tucker recompression pipeline are proposed as direct corollaries of our structural theory. Numerical experiments across three distinct families of high-dimensional test functions validate all theoretical predictions and confirm the intrinsic low-rank compressibility of hyperinterpolation coefficients. In contrast to prior algorithm-centric studies, this work prioritizes rigorous theoretical characterization over implementation tricks, establishing a unified structural and mathematical foundation for high-dimensional hyperinterpolation.

math.NA↗

arXiv2026

Speculative Correction: Draft-then-Refine Decoding for Diffusion Language Models

Diffusion language models (DLMs) can revise tokens bidirectionally, but standard decoding procedures often adapt them to left-to-right generation by producing text block by block. We study a simple plug-and-play inference pattern: first generate a complete draft, then refine the full response using bidirectional diffusion. Using LLaDA2.1-Flash and LLaDA2.1-Mini, we evaluate two configurations. In Flash-Flash, the same Flash model serves as both drafter and refiner, testing whether an existing model can improve its own block-autoregressive output through global refinement. In Mini-Flash, inspired by speculative decoding, we introduce speculative correction: Mini drafts a full response, and Flash revises it as an editable initialization. Flash-Flash improves GSM8K-384 accuracy from 0.848 to 0.899 while running 1.20 times faster than the selected Flash block-autoregressive baseline, and improves MBPP-384 from 0.545 to 0.693. Latency-window-matched Flash-only controls indicate that these gains persist after targeted tuning of block-autoregressive decoding. Causal ablations indicate that completed drafts provide useful initializations: refinement from a fully masked span performs poorly, full global refinement provides a clear additional gain on GSM8K, and local refinement captures much of the gain on MBPP and MATH. Mini-Flash provides useful quality-latency trade-offs, including MATH-384 performance of 0.294 versus 0.300 for Flash while running 2.17 times faster. These results support a Pareto-frontier interpretation rather than the claim that the heterogeneous cascade uniformly matches Flash quality. Overall, same-model draft-and-refine provides evidence that bidirectional refinement is a useful decoding primitive for DLMs, while speculative correction demonstrates a training-free route to fast DLM generation.

cs.CL↗

arXiv2026

Statistical Proof as a Window into Human-AI Collaboration: Practical Insights and a Community Agenda

Large language models (LLMs) are increasingly woven into expert cognitive work in daily research, yet we know little about how human expertise should adapt when an AI system can execute substantial technical reasoning on its own. Here we use statistical proof development, a demanding and structured form of expert reasoning, as a window into this broader question. Drawing on day-to-day proof problems, we find that current general-purpose LLMs occupy a useful but limited role: they can execute technical components given a precisely formulated problem and targeted guidance, but become unreliable when the problem is open-ended or requires a long reasoning chain with multiple interdependent steps. This execution-strategy gap is rooted in what makes research-level statistical proof distinctive: unlike pure mathematics, where problems arrive pre-formulated and often demand novel techniques, statistical proof requires first modeling a scientific question into a statistical framework with appropriate assumptions, and then identifying and adapting the right strategy from a repertoire of reusable domain-specific tools. Each step requires deep expertise in both the statistical literature and the real-world context being modeled. In such work, current AI assistance does not reduce the need for human expertise; it relocates that expertise to where human decision-making matters most, such as problem formulation and verification of AI-generated results, and may raise the bar for both. These findings yield practical suggestions for how statisticians can structure AI-assisted proof workflows, and point to a broader community agenda for shared resources, better AI tools, and training the next generation of researchers. Using statistical proof as a window, our study has implications for how experts structure human-AI collaboration in technical cognitive domains more broadly.

stat.OT↗

arXiv2026

BFLA: Block-Filtered Long-Context Attention Mechanism

This paper proposes Block-Filtered Long-Context Attention (BFLA), a training-free sparse prefill attention mechanism for long-context inference. BFLA adopts a two-stage design. In Stage 1, query and key sequences are compressed into coarse blocks, and lightweight block-level softmax mass estimation is performed to construct an input-dependent block importance mask. In Stage 2, the coarse mask is expanded to the Triton attention-tile grid. Several tile-level rescue strategies are applied to reduce information loss, where a fused sparse prefill kernel skips unimportant KV tiles while preserving exact token-level attention inside every retained tile. BFLA requires no retraining, calibration, preprocessing, or model modification and can be plugged into existing vLLM-style paged-attention workloads. Experiments on Gemma 4, Llama 3.1, Qwen 3.5, and Qwen 3.6 series models show that BFLA substantially accelerates long-context prefilling with minimal accuracy degradation compared to dense Triton FlashAttention. Project website: https://github.com/Alicewithrabbit/BFLA.

eess.SP↗

arXiv2025

UniFormer: Unified and Efficient Transformer for Reasoning Across General and Custom Computing

The success of neural networks such as convolutional neural networks (CNNs) has been largely attributed to their effective and widespread deployment on customised computing platforms, including field-programmable gate arrays (FPGAs) and application-specific integrated circuits (ASICs). In the current era, Transformer-based architectures underpin the majority of state-of-the-art (SOTA) larger models that are also increasingly deployed on customised computing hardware for low-power and real-time applications. However, the fundamentally different parallel computation paradigms between general-purpose and customised computing often lead to compromises in model transfer and deployability, which typically come at the cost of complexity, efficiency or accuracy. Moreover, many cross-platform optimisation principles have also remained underexplored in existing studies. This paper introduces UniFormer, a unified and efficient Transformer architecture for both general-purpose and customised computing platforms. By enabling higher parallelism and compute-storage fusion, UniFormer achieves state-of-the-art (SOTA) accuracy and latency on GPUs while exhibiting strong adaptability on FPGAs. To the best of our knowledge, this paper is the first efficient Transformer work that jointly considers both general-purpose and customised computing architectures.

cs.DC↗

arXiv2025

Meta-Router: Bridging Gold-standard and Preference-based Evaluations in Large Language Model Routing

In language tasks that require extensive human--model interaction, deploying a single "best" model for every query can be expensive. To reduce inference cost while preserving the quality of the responses, a large language model (LLM) router selects the most appropriate model from a pool of candidates for each query. A central challenge to training a high-quality router is the scarcity of reliable supervision. Gold-standard data (e.g., expert-verified labels or rubric-based scores) provide accurate quality evaluations of LLM responses but are costly and difficult to scale. In contrast, preference-based data, collected via crowdsourcing or LLM-as-a-judge systems, are cheaper and more scalable, yet often biased in reflecting the true quality of responses. We cast the problem of LLM router training with combined gold-standard and preference-based data into a causal inference framework by viewing the response evaluation mechanism as the treatment assignment. This perspective further reveals that the bias in preference-based data corresponds to the well-known causal estimand: the conditional average treatment effect. Based on this new perspective, we develop an integrative causal router training framework that corrects preference-data bias, address imbalances between two data sources, and improve routing robustness and efficiency. Numerical experiments demonstrate that our approach delivers more accurate routing and improves the trade-off between cost and quality.

cs.LG↗

arXiv2025

sparseGeoHOPCA: A Geometric Solution to Sparse Higher-Order PCA Without Covariance Estimation

We propose sparseGeoHOPCA, a novel framework for sparse higher-order principal component analysis (SHOPCA) that introduces a geometric perspective to high-dimensional tensor decomposition. By unfolding the input tensor along each mode and reformulating the resulting subproblems as structured binary linear optimization problems, our method transforms the original nonconvex sparse objective into a tractable geometric form. This eliminates the need for explicit covariance estimation and iterative deflation, enabling significant gains in both computational efficiency and interpretability, particularly in high-dimensional and unbalanced data scenarios. We theoretically establish the equivalence between the geometric subproblems and the original SHOPCA formulation, and derive worst-case approximation error bounds based on classical PCA residuals, providing data-dependent performance guarantees. The proposed algorithm achieves a total computational complexity of $O\left(\sum_{n=1}^{N} (k_n^3 + J_n k_n^2)\right)$, which scales linearly with tensor size. Extensive experiments demonstrate that sparseGeoHOPCA accurately recovers sparse supports in synthetic settings, preserves classification performance under 10$\times$ compression, and achieves high-quality image reconstruction on ImageNet, highlighting its robustness and versatility.

math.NA↗

arXiv2025

Efficient randomized algorithms for the fixed Tucker-rank problem of Tucker decomposition with adaptive shifts

Randomized numerical linear algebra is proved to bridge theoretical advancements to offer scalable solutions for approximating tensor decomposition. This paper introduces fast randomized algorithms for solving the fixed Tucker-rank problem of Tucker decomposition, through the integration of adaptive shifted power iterations. The proposed algorithms enhance randomized variants of truncated high-order singular value decomposition (T-HOSVD) and sequentially T-HOSVD (ST-HOSVD) by incorporating dynamic shift strategies, which accelerate convergence by refining the singular value gap and reduce the number of required power iterations while maintaining accuracy. Theoretical analyses provide probabilistic error bounds, demonstrating that the proposed methods achieve comparable or superior accuracy compared to deterministic approaches. Numerical experiments on synthetic and real-world datasets validate the efficiency and robustness of the proposed algorithms, showing a significant decline in runtime and approximation error over state-of-the-art techniques.

math.NA↗

arXiv2025

RSVP: Reasoning Segmentation via Visual Prompting and Multi-modal Chain-of-Thought

Multi-modal Large Language Models (MLLMs) have demonstrated remarkable reasoning capability while lack explicit mechanisms for visual grounding and segmentation, creating a gap between cognitive reasoning and visual perception. To bridge this gap, we introduce Reasoning Segmentation via Visual Prompting (RSVP), a novel framework that unifies multi-step multimodal reasoning with grounded visual understanding. RSVP is a two-stage structuralized framework that integrates reasoning-driven localization with segmentation refinement. In the reasoning stage, RSVP employs multimodal chain-of-thought visual prompts to help MLLMs understand queries and infer targets, generating interpretable region proposals that enhance visual grounding. In segmentation stage, RSVP refines these proposals with a Vision-Language Segmentation Module (VLSM), seamlessly integrates textual and visual cues to produce precise segmentation masks. By explicitly modelling the interaction between multimodal reasoning and segmentation, RSVP introduces a new paradigm for interpretable reasoning segmentation. It exploits MLLMs' inherent localization capabilities, enabling the models to not only reason about objects but also generate structured visual representations. Our extensive experiments demonstrate that RSVP achieves state-of-the-art performance, surpasses state-of-the-art methods by up to +6.5 gIoU and +9.2 cIoU on ReasonSeg, and achieves 49.7 mAP on SegInW under zero-shot settings. These results validate RSVP as an effective and scalable framework for integrating cognitive reasoning with structured visual understanding.

cs.CV↗

arXiv2025

ELFATT: Efficient Linear Fast Attention for Vision Transformers

The attention mechanism is the key to the success of transformers in different machine learning tasks. However, the quadratic complexity with respect to the sequence length of the vanilla softmax-based attention mechanism becomes the major bottleneck for the application of long sequence tasks, such as vision tasks. Although various efficient linear attention mechanisms have been proposed, they need to sacrifice performance to achieve high efficiency. What's more, memory-efficient methods, such as FlashAttention-1-3, still have quadratic computation complexity which can be further improved. In this paper, we propose a novel efficient linear fast attention (ELFATT) mechanism to achieve low memory input/output operations, linear computational complexity, and high performance at the same time. ELFATT offers 4-7x speedups over the vanilla softmax-based attention mechanism in high-resolution vision tasks without losing performance. ELFATT is FlashAttention friendly. Using FlashAttention-2 acceleration, ELFATT still offers 2-3x speedups over the vanilla softmax-based attention mechanism on high-resolution vision tasks without losing performance. Even in some non-vision tasks of long-range arena, ELFATT still achieves leading performance and offers 1.2-2.3x speedups over FlashAttention-2. Even on edge GPUs, ELFATT still offers 1.6x to 2.0x speedups compared to state-of-the-art attention mechanisms in various power modes from 5W to 60W. Furthermore, ELFATT can be used to enhance and accelerate diffusion tasks directly without training.

eess.IV↗

arXiv2024

Local Balance Calibration for Nonparametric Propensity Score Estimation

The propensity score is widely used for causal inference in observational studies, but common parametric estimators can produce biased and inefficient effect estimates when model assumptions are violated. Nonparametric approaches reduce sensitivity to misspecification but often yield unstable weights and inadequate covariate balance. We propose Local Balance with Calibration, implemented by Neural Networks, a weighting method that combines flexible function approximation with the explicit enforcement of covariate balance and calibration. When used with inverse probability weighting, the proposed estimator produces more stable weights, improved covariate balance, and reduced bias in average treatment effect estimation compared with existing approaches. We further develop an influence-function-based variance estimator that provides accurate uncertainty quantification for the resulting weighted estimators. Numerical studies demonstrate improved efficiency and reliable variance estimation across a range of data-generating scenarios. The method is implemented using the publicly available R package LBCNet.

stat.ME↗

arXiv2023

Mediation Analysis with Mendelian Randomization and Efficient Multiple GWAS Integration

Mediation analysis is a powerful tool for studying causal pathways between exposure, mediator, and outcome variables of interest. While classical mediation analysis using observational data often requires strong and sometimes unrealistic assumptions, such as unconfoundedness, Mendelian Randomization (MR) avoids unmeasured confounding bias by employing genetic variations as instrumental variables. We develop a novel MR framework for mediation analysis with genome-wide associate study (GWAS) summary data, and provide solid statistical guarantees. Our framework employs carefully crafted estimating equations, allowing for different sets of genetic variations to instrument the exposure and the mediator, to efficiently integrate information stored in three independent GWAS. As part of this endeavor, we demonstrate that in mediation analysis, the challenge raised by instrument selection goes beyond the well-known winner's curse issue, and therefore, addressing it requires special treatment. We then develop bias correction techniques to address the instrument selection issue and commonly encountered measurement error bias issue. Collectively, through our theoretical investigations, we show that our framework provides valid statistical inference for both direct and mediation effects with enhanced statistical efficiency compared to existing methods. We further illustrate the finite-sample performance of our approach through simulation experiments and a case study.

stat.ME↗

arXiv2023

Multi-View Variational Autoencoder for Missing Value Imputation in Untargeted Metabolomics

Background: Missing data is a common challenge in mass spectrometry-based metabolomics, which can lead to biased and incomplete analyses. The integration of whole-genome sequencing (WGS) data with metabolomics data has emerged as a promising approach to enhance the accuracy of data imputation in metabolomics studies. Method: In this study, we propose a novel method that leverages the information from WGS data and reference metabolites to impute unknown metabolites. Our approach utilizes a multi-view variational autoencoder to jointly model the burden score, polygenetic risk score (PGS), and linkage disequilibrium (LD) pruned single nucleotide polymorphisms (SNPs) for feature extraction and missing metabolomics data imputation. By learning the latent representations of both omics data, our method can effectively impute missing metabolomics values based on genomic information. Results: We evaluate the performance of our method on empirical metabolomics datasets with missing values and demonstrate its superiority compared to conventional imputation techniques. Using 35 template metabolites derived burden scores, PGS and LD-pruned SNPs, the proposed methods achieved R^2-scores > 0.01 for 71.55% of metabolites. Conclusion: The integration of WGS data in metabolomics imputation not only improves data completeness but also enhances downstream analyses, paving the way for more comprehensive and accurate investigations of metabolic pathways and disease associations. Our findings offer valuable insights into the potential benefits of utilizing WGS data for metabolomics data imputation and underscore the importance of leveraging multi-modal data integration in precision medicine research.

q-bio.GN↗

arXiv2023

Physics Inspired Hybrid Attention for SAR Target Recognition

There has been a recent emphasis on integrating physical models and deep neural networks (DNNs) for SAR target recognition, to improve performance and achieve a higher level of physical interpretability. The attributed scattering center (ASC) parameters garnered the most interest, being considered as additional input data or features for fusion in most methods. However, the performance greatly depends on the ASC optimization result, and the fusion strategy is not adaptable to different types of physical information. Meanwhile, the current evaluation scheme is inadequate to assess the model's robustness and generalizability. Thus, we propose a physics inspired hybrid attention (PIHA) mechanism and the once-for-all (OFA) evaluation protocol to address the above issues. PIHA leverages the high-level semantics of physical information to activate and guide the feature group aware of local semantics of target, so as to re-weight the feature importance based on knowledge prior. It is flexible and generally applicable to various physical models, and can be integrated into arbitrary DNNs without modifying the original architecture. The experiments involve a rigorous assessment using the proposed OFA, which entails training and validating a model on either sufficient or limited data and evaluating on multiple test sets with different data distributions. Our method outperforms other state-of-the-art approaches in 12 test scenarios with same ASC parameters. Moreover, we analyze the working mechanism of PIHA and evaluate various PIHA enabled DNNs. The experiments also show PIHA is effective for different physical information. The source code together with the adopted physical information is available at https://github.com/XAI4SAR.

cs.CV↗

arXiv2023

Winner's Curse Free Robust Mendelian Randomization with Summary Data

In the past decade, the increased availability of genome-wide association studies summary data has popularized Mendelian Randomization (MR) for conducting causal inference. MR analyses, incorporating genetic variants as instrumental variables, are known for their robustness against reverse causation bias and unmeasured confounders. Nevertheless, classical MR analyses utilizing summary data may still produce biased causal effect estimates due to the winner's curse and pleiotropic issues. To address these two issues and establish valid causal conclusions, we propose a unified robust Mendelian Randomization framework with summary data, which systematically removes the winner's curse and screens out invalid genetic instruments with pleiotropic effects. Different from existing robust MR literature, our framework delivers valid statistical inference on the causal effect neither requiring the genetic pleiotropy effects to follow any parametric distribution nor relying on perfect instrument screening property. Under appropriate conditions, we show that our proposed estimator converges to a normal distribution and its variance can be well estimated. We demonstrate the performance of our proposed estimator through Monte Carlo simulations and two case studies. The codes implementing the procedures are available at https://github.com/ChongWuLab/CARE/.

stat.ME↗

arXiv2023

Breaking the Winner's Curse in Mendelian Randomization: Rerandomized Inverse Variance Weighted Estimator

Developments in genome-wide association studies and the increasing availability of summary genetic association data have made the application of two-sample Mendelian Randomization (MR) with summary data increasingly popular. Conventional two-sample MR methods often employ the same sample for selecting relevant genetic variants and for constructing final causal estimates. Such a practice often leads to biased causal effect estimates due to the well known "winner's curse" phenomenon. To address this fundamental challenge, we first examine its consequence on causal effect estimation both theoretically and empirically. We then propose a novel framework that systematically breaks the winner's curse, leading to unbiased association effect estimates for the selected genetic variants. Building upon the proposed framework, we introduce a novel rerandomized inverse variance weighted estimator that is consistent when selection and parameter estimation are conducted on the same sample. Under appropriate conditions, we show that the proposed RIVW estimator for the causal effect converges to a normal distribution asymptotically and its variance can be well estimated. We illustrate the finite-sample performance of our approach through Monte Carlo experiments and two empirical examples.

stat.ME↗

arXiv2022

Efficient Targeted Learning of Heterogeneous Treatment Effects for Multiple Subgroups

In biomedical science, analyzing treatment effect heterogeneity plays an essential role in assisting personalized medicine. The main goals of analyzing treatment effect heterogeneity include estimating treatment effects in clinically relevant subgroups and predicting whether a patient subpopulation might benefit from a particular treatment. Conventional approaches often evaluate the subgroup treatment effects via parametric modeling and can thus be susceptible to model mis-specifications. In this manuscript, we take a model-free semiparametric perspective and aim to efficiently evaluate the heterogeneous treatment effects of multiple subgroups simultaneously under the one-step targeted maximum-likelihood estimation (TMLE) framework. When the number of subgroups is large, we further expand this path of research by looking at a variation of the one-step TMLE that is robust to the presence of small estimated propensity scores in finite samples. From our simulations, our method demonstrates substantial finite sample improvements compared to conventional methods. In a case study, our method unveils the potential treatment effect heterogeneity of rs12916-T allele (a proxy for statin usage) in decreasing Alzheimer's disease risk.

stat.ME↗

arXiv2022

Leveraging Graph-based Cross-modal Information Fusion for Neural Sign Language Translation

Sign Language (SL), as the mother tongue of the deaf community, is a special visual language that most hearing people cannot understand. In recent years, neural Sign Language Translation (SLT), as a possible way for bridging communication gap between the deaf and the hearing people, has attracted widespread academic attention. We found that the current mainstream end-to-end neural SLT models, which tries to learning language knowledge in a weakly supervised manner, could not mine enough semantic information under the condition of low data resources. Therefore, we propose to introduce additional word-level semantic knowledge of sign language linguistics to assist in improving current end-to-end neural SLT models. Concretely, we propose a novel neural SLT model with multi-modal feature fusion based on the dynamic graph, in which the cross-modal information, i.e. text and video, is first assembled as a dynamic graph according to their correlation, and then the graph is processed by a multi-modal graph encoder to generate the multi-modal embeddings for further usage in the subsequent neural translation models. To the best of our knowledge, we are the first to introduce graph neural networks, for fusing multi-modal information, into neural sign language translation models. Moreover, we conducted experiments on a publicly available popular SLT dataset RWTH-PHOENIX-Weather-2014T. and the quantitative experiments show that our method can improve the model.

cs.CL↗