arXiv subjects

Kexin Wang

Publications and source records attributed to Kexin Wang.

At least 19 recordsLinked to original sources

arXiv2026

Beyond Transformers: Linear Attention Policy for Open-Vocabulary Object Goal Navigation

Open-Vocabulary Object Goal Navigation (OVON) requires agents to operate under partial observability, making effective internal state updates critical for navigation performance. This update is implemented by the policy network, where recent approaches adopt Transformer-based backbones with self-attention over a context window to integrate temporal information. However, our controlled experiments show that performance does not scale with context length under Transformer-based policies, questioning the suitability of self-attention for state integration in navigation. To this end, we propose Linear Attention-based Navigation (LANav), which adopts linear attention (LA) as the policy backbone to maintain a structured state update rather than self-attention over the context window. Across multiple LA variants evaluated under identical settings, LANav consistently outperforms Transformer-based baselines. Performance improves as state update mechanisms become more structured and regulated, highlighting the importance of state update design. To improve state update effectiveness, we introduce Weighted State-Expansion Linear Attention (WSLA), which expands each attention head's state into multiple sub-states and uses learnable weighted readout to aggregate expanded sub-states. Equipped with WSLA, LANav achieves 36.4% average success rate (SR) on HM3D-OVON, outperforming Transformer-based counterparts by 6.3 percentage points in macro-averaged SR, while maintaining computational efficiency. Distance-stratified results show larger gains in long-distance episodes, while HSSD transfer and fine-tuning demonstrate robustness across scene distributions. Real-world deployment on a Unitree Go2 further achieves an 82% success rate over 50 trials, supporting the practical feasibility and sim-to-real transfer of LANav.

cs.RO↗

arXiv2026

Electron Delocalization versus Emission Coherence of Quantum Dot Superlattices

Cooperative emission is a collective quantum optical process that requires macroscopic phase coherence among coupled emitters. Recent observations of cooperative emission in QD superlattices have renewed interest in how such coherence emerges in nanostructured solids. Meanwhile, theoretical studies have long discussed the relationship between electronic delocalization and coherence, particularly whether delocalized states necessarily give rise to cooperative emission. This study addresses this question through power-dependent steady-state PL and time-resolved PL decay measurements. The findings indicate that, although the quantum resonance peak exhibits delocalized excitonic characteristics, it shows no signatures of cooperative radiation. In particular, neither superlinear intensity scaling nor power-dependent emission delay was observed, indicating the absence of cooperative-radiation signatures. This can be understood from two disorder-related aspects. Temperature-dependent spectroscopy reveals pronounced inhomogeneous broadening and low-temperature dark-exciton participation, pointing to intra-domain static disorder and exciton-state mixing. These effects collectively hinder the establishment of macroscopic coherence. The temperature dependence of the quantum resonance peak decay lifetime is consistent with two-dimensional exciton dynamics. This work provides direct experimental evidence that electronic delocalization can be decoupled from cooperative coherence in CdSe quantum dot superlattices.

cond-mat.mes-hall↗

arXiv2026

Tensor-based second-order causal discovery

Causal discovery seeks to uncover the causal dependencies among variables. For this purpose, we propose an algorithm called Tensor-based Second-order Causal Discovery (TSCD). Its input is a tensor obtained from the covariance matrices of observational and interventional data. Assuming the causal dependencies follow a linear structural equation model on a directed acyclic graph (DAG), TSCD outputs the DAG and the functions on its edges, requiring only that the noise variables are uncorrelated. We also implement a version of the approach for nonlinear models. Our focus on second-order statistics (via the covariance matrices) is motivated by their statistical and computational efficiency relative to higher-order moments, their identifiability relative to first-order statistics, and that they work regardless of whether the variables are Gaussian. We show that TSCD has identifiable causal order and parameters from a number of interventions that is logarithmic in the number of variables. Experiments show that TSCD is robust to noise, competitive with existing methods, and scales to hundreds of variables.

stat.ML↗

arXiv2026

HyNeuralMap: Hyperbolic Mapping of Visual Semantics to Neural Hierarchies

Understanding the intricate mappings between visual stimuli and neural responses is a fundamental challenge in cognitive neuroscience. While current approaches predominantly align images and functional magnetic resonance imaging (fMRI) responses in Euclidean space, this geometry often struggles to preserve fine-grained semantic relationships and latent hierarchical structures across visual and neural modalities. To overcome this, we propose HyNeuralMap, a framework that employ hyperbolic Lorentz model to map visual semantics into a shared, cross-subject neural hierarchy. By leveraging the negative curvature of hyperbolic space as an inductive bias, the proposed framework better captures hierarchical semantic organization and cross-subject neural similarities. Specifically, visual and neural embeddings are jointly optimized through hyperbolic geometric alignment, where geodesic distances preserve semantic proximity and hierarchical relationships more effectively than Euclidean embeddings. Experiments demonstrate that HyNeuralMap consistently outperforms state-of-the-art Euclidean baselines in both multi-label semantic prediction and cross-modal retrieval tasks. This confirms hyperbolic geometry's superiority for cross-modal semantic alignment and hierarchical modeling, providing a new avenue for vision-neural representation learning.

cs.CV↗

arXiv2026

When Altruism Meets Autonomy: Managing Bottleneck Congestion with Strategic Autonomous Vehicles

Weaving ramps are critical bottlenecks in highway networks due to conflicting traffic flows and complex interactions among heterogeneous vehicle types. In mixed-autonomy settings, the presence of controllable autonomous vehicles (AVs) introduces new opportunities to influence system-level outcomes, yet the structural impact of such control remains poorly understood. This paper develops a unified equilibrium framework to capture, predict, and optimize aggregate lane-choice behavior in weaving ramps with heterogeneous vehicle populations. We first formulate a Wardrop-based model capturing the selfish behavior of human-driven vehicles (HDVs) and establish existence, uniqueness, and validity of the resulting equilibrium. We then introduce a Stackelberg--Wardrop formulation in which AVs act as strategic leaders optimizing system performance, while HDVs respond through equilibrium adaptation. The framework is further generalized to incorporate heterogeneous behavioral preferences of HDVs and AVs via a Social Value Orientation (SVO) model. Our analysis reveals a fundamental structural property of mixed-autonomy traffic systems: under selfish HDV behavior, the impact of AV penetration is inherently non-increasing, exhibiting plateau regions where performance remains unchanged and improves only at critical thresholds. These results provide principled guidance for the design of AV control and incentive mechanisms in the presence of selfish human behavior, and demonstrate how strategically controlled autonomous agents can be deployed to induce system-level efficiency gains in mixed-autonomy transportation networks.

eess.SY↗

arXiv2026

Masked Training for Robust Arrhythmia Detection from Digitalized Multiple Layout ECG Images

Background: Electrocardiograms are indispensable for diagnosing cardiovascular diseases, yet in many settings they exist only as paper printouts stored in multiple recording layouts. Converting these images into digital signals introduces two key challenges: temporal asynchrony among leads and partial blackout missing, where contiguous signal segments become entirely unavailable. Existing models cannot adequately handle these concurrent problems while maintaining interpretability. Methods: We propose PatchECG, combining an adaptive variable block count missing learning mechanism with a masked training strategy. The model segments each lead into fixed-length patches, discards entirely missing patches, and encodes the remainder via a pluggable patch encoder. A disordered patch attention mechanism with patch-level temporal and lead embeddings captures cross-lead and temporal dependencies without interpolation. PatchECG was trained on PTB-XL and evaluated under seven simulated layout conditions, with external validation on 400 real ECG images from Chaoyang Hospital across three clinical layouts. Results: PatchECG achieves an average AUROC of approximately 0.835 across all simulated layouts. On the Chaoyang cohort, the model attains an overall AUROC of 0.778 for atrial fibrillation detection, rising to 0.893 on the 12x1 subset -- surpassing the pre-trained baseline by 0.111 and 0.190, respectively. Model attention aligns with cardiologist annotations at a rate approaching inter-clinician agreement. Conclusions: PatchECG provides a robust, interpolation-free, and interpretable solution for arrhythmia detection from digitized ECG images across diverse layouts. Its direct modeling of asynchronous and partially missing signals, combined with clinically aligned attention, positions it as a practical tool for cardiac diagnostics from legacy ECG archives in real-world clinical environments.

cs.LG↗

arXiv2026

Piecewise M-Stationarity and Related Algorithms for Mathematical Programs with Complementarity Constraints

This study explores B-stationarity of mathematical programs with complementarity constraints (MPCCs) and convergence behavior of MPCC algorithms. Special attention is given to the cases with biactive complementarity constraints. First, we propose the concept of piecewise M-stationarity and prove its equivalence to B-stationarity under MPCC-GCQ. Then, we investigate convergence properties of the NCP-based bounding methods we proposed in [31], by requiring MPCC-MFCQ; an interpretation of the algorithm's behavior together with the concept of piecewise M-stationarity leads to a cost reduction in B-stationarity verification. In addition, practical issues related to convergence to non-strongly stationary solutions are discussed, which shows that the NCP-based complementarity reformulations have an advantage in avoiding unbounded multipliers near these solutions.

math.OC↗

arXiv2026

Electrically tunable circular photocurrent via local-field induced symmetry breaking at a metal-MoTe2 interface

Transition metal dichalcogenides (TMDCs) constitute a promising platform for symmetry-engineered responses to circularly polarized light. The high crystal symmetry of centrosymmetric 2H-phase TMDCs inherently forbids the circular photogalvanic effect, thereby necessitating external stimuli such as electric fields or strain to lower the symmetry for its activation. While Schottky junctions provide a ubiquitous built-in field for potentially inducing circular photocurrents, the mechanism for the generation and control of circular photocurrents in TMDCs is not understood. In this study, we fabricated a localized gold-MoTe2 heterostructure and demonstrate a pronounced circular photocurrent at the interface under normal incidence. The photocurrent is attributed to circular photogalvanic effect governed by the strength and direction of the built-in electric field, enabling continuous modulation via an external bias. First-principles calculations show that the gold interface induces a spin splitting in the valence bands of MoTe2, establishing a valley-dependent spin ordering. The observed circular photocurrent from multilayer 2H-MoTe2 under normal incidence indicates the breaking of C3 rotational symmetry by the local in-plane field. These results establish an effective strategy for developing voltage-tunable circularly polarized photodetectors and valleytronic devices.

cond-mat.mtrl-sci↗

arXiv2026

A Real Generalized Trisecant Trichotomy

The classical trisecant lemma says that a general chord of a non-degenerate space curve is not a trisecant; that is, the chord only meets the curve in two points. The generalized trisecant lemma extends the result to higher-dimensional varieties. It states that the linear space spanned by general points on a projective variety intersects the variety in exactly these points, provided the dimension of the linear space is smaller than the codimension of the variety and that the variety is irreducible, reduced, and non-degenerate. We prove a real analogue of the generalized trisecant lemma, which takes the form of a trichotomy. Along the way, we characterize the possible numbers of real intersection points between a real projective variety and a complimentary dimension real linear space. We show that any integer of correct parity between a minimum and a maximum number can be achieved. We then specialize to Segre-Veronese varieties, where our results apply to the identifiability of independent component analysis, tensor decomposition and to typical tensor ranks.

math.AG↗

arXiv2026

Contrastive independent component analysis

In recent years, there has been growing interest in jointly analyzing a foreground dataset, representing an experimental group, and a background dataset, representing a control group. The goal of such contrastive investigations is to identify salient features in the experimental group relative to the control. Independent component analysis (ICA) is a powerful tool for learning independent patterns in a dataset. We generalize it to contrastive ICA (cICA). For this purpose, we devise a new linear algebra based tensor decomposition algorithm, which is more expressive but just as efficient and identifiable as other linear algebra based algorithms. We establish the identifiability of cICA and demonstrate its performance in finding patterns and visualizing data, using synthetic, semi-synthetic, and real-world datasets, comparing the approach to existing methods.

math.ST↗

arXiv2026

Multi-context principal component analysis

Principal component analysis (PCA) is a tool to capture factors that explain variation in data. Across domains, data are now collected across multiple contexts (for example, individuals with different diseases, cells of different types, or words across texts). While the factors explaining variation in data are undoubtedly shared across subsets of contexts, no tools currently exist to systematically recover such factors. We develop multi-context principal component analysis (MCPCA), a theoretical and algorithmic framework that decomposes data into factors shared across subsets of contexts. Applied to gene expression, MCPCA reveals axes of variation shared across subsets of cancer types and an axis whose variability in tumor cells, but not mean, is associated with lung cancer progression. Applied to contextualized word embeddings from language models, MCPCA maps stages of a debate on human nature, revealing a discussion between science and fiction over decades. These axes are not found by combining data across contexts or by restricting to individual contexts. MCPCA is a principled generalization of PCA to address the challenge of understanding factors underlying data across contexts.

stat.ML↗

arXiv2026

A Curriculum Learning Approach to Reinforcement Learning: Leveraging RAG for Multimodal Question Answering

This paper describes the solutions of the Dianping-Trust-Safety team for the META CRAG-MM challenge. The challenge requires building a comprehensive retrieval-augmented generation system capable for multi-modal multi-turn question answering. The competition consists of three tasks: (1) answering questions using structured data retrieved from an image-based mock knowledge graph, (2) synthesizing information from both knowledge graphs and web search results, and (3) handling multi-turn conversations that require context understanding and information aggregation from multiple sources. For Task 1, our solution is based on the vision large language model, enhanced by supervised fine-tuning with knowledge distilled from GPT-4.1. We further applied curriculum learning strategies to guide reinforcement learning, resulting in improved answer accuracy and reduced hallucination. For Task 2 and Task 3, we additionally leveraged web search APIs to incorporate external knowledge, enabling the system to better handle complex queries and multi-turn conversations. Our approach achieved 1st place in Task 1 with a significant lead of 52.38%, and 3rd place in Task 3, demonstrating the effectiveness of the integration of curriculum learning with reinforcement learning in our training pipeline.

cs.AI↗

arXiv2026

Physics-embedded neural computational electron microscopy for quantitative 4D nanometrology

The fusion of rigorous physical laws with flexible data-driven learning represents a new frontier in scientific simulation, yet bridging the gap between physical interpretability and computational efficiency remains a grand challenge. In electron microscopy, this divide limits the ability to quantify three-dimensional topography from two-dimensional projections, fundamentally constraining our understanding of nanoscale structure-function relationships. Here, we present a physics-embedded neural computational microscopy framework that achieves metrological three-dimensional reconstruction by deeply coupling a differentiable electron-optical forward model with deep learning. By introducing a Vision Field Transformer as a high-speed, differentiable surrogate for physical process analysis simulations, we establish an end-to-end, self-supervised optimization loop that enforces strict physical consistency with hardware geometry. This synergy enables single-shot, quantitative three-dimensional nanometrology with precision comparable to atomic force microscopy but at orders of magnitude higher throughput. Furthermore, we demonstrate the capability for four-dimensional (3D real space plus time) in situ characterization by tracking the dynamic evolution of surface nanostructure during copper redox, revealing hidden crystallographic kinetics invisible to conventional imaging. Our work not only redefines the limits of scanning electron microscopy but also establishes a generalizable archetype for solving ill-posed inverse problems across physical sciences, unlocking the full potential of simulation as a third pillar of discovery.

physics.optics↗

arXiv2025

Traffic Equilibrium in Mixed-Autonomy Network with Capped Customer Waiting

This paper develops a unified modeling framework to capture the equilibrium-state interactions among ride-hailing companies, travelers, and traffic of mixed-autonomy transportation networks. Our framework integrates four interrelated sub-modules: (i) the operational behavior of representative ride-hailing Mixed-Fleet Traffic Network Companies (MiFleet TNCs) managing autonomous vehicle (AV) and human-driven vehicle (HV) fleets, (ii) traveler mode-choice decisions taking into account travel costs and waiting time, (iii) capped customer waiting times to reflect the option available to travelers not to wait for TNCs' service beyond his/her patience and to resort to existing travel modes, and (iv) a flow-dependent traffic congestion model for travel times. A key modeling feature distinguishes AVs and HVs across the pickup and service (customer-on-board) stages: AVs follow Wardrop pickup routes but may deviate during service under company coordination, whereas HVs operate in the reverse manner. The overall framework is formulated as a Nonlinear Complementarity Problem (NCP), which is equivalent to a Variational Inequality(VI) formulation based on which the existence of a variational equilibrium solution to the traffic model is established. Numerical experiments examine how AV penetration and Wardrop relaxation factors, which bound route deviation, affect company, traveler, and system performance to various degrees. The results provide actionable insights for policymakers on regulating AV adoption and company vehicle deviation behavior in modern-day traffic systems that are fast changing due to the advances in technology and information accessibility.

eess.SY↗

arXiv2025

A Hands-On Workshop for Constructing a Low-Field MRI System in Three Days

Access to Magnetic Resonance Imaging system assembly knowledge can be expanded by leveraging open-source hardware and software, simplified installation requirements, and collaborative training initiatives. To this end, we conducted a three-day workshop to construct an operational 0.27T MRI scanner. The workshop hosted 16 participants, including faculty, postdoctoral fellows, trainers, and students, who collaborated to build the scanner using open-source hardware and software components. Teams were designated to focus on various subsystems, including the magnet, passive shimming, radiofrequency (RF) coils, gradient coils, data acquisition, and reconstruction. Pre-workshop preparation involved simulation-based design processes and fabrication techniques, which incorporated configuring MaRCoS and PyPulseq libraries, CNC machining, and 3D printing. During the workshop, participants assembled an H-shaped magnet, which achieved a peak magnetic field strength of 0.269T. Passive shimming effectively reduced the field inhomogeneity from 3mT to 2mT. A 3 cm diameter RF solenoid was built and tuned to 11.4 MHz. The gradients exhibited less than 5% non-linearity in simulations and were fabricated by CNC machining copper plates. The assembled system was used to acquire a 2D spin echo of a water phantom. Following the workshop, the system was further optimized to scan relaxometry phantoms. A post-workshop survey was carried out, revealing over 87% satisfaction. The constructed scanner represents a valuable platform for educational initiatives, pulse sequence development, and preclinical research imaging efforts.

physics.ins-det↗

arXiv2025

Portfolio Optimization via Transfer Learning

Recognizing that asset markets generally exhibit shared informational characteristics, we develop a portfolio strategy based on transfer learning that leverages cross-market information to enhance the investment performance in the market of interest by forward validation. Our strategy asymptotically identifies and utilizes the informative datasets, selectively incorporating valid information while discarding the misleading information. This enables our strategy to achieve the maximum Sharpe ratio asymptotically. The promising performance is demonstrated by numerical studies and case studies of two portfolios: one consisting of stocks dual-listed in A-shares and H-shares, and another comprising equities from various industries of the United States.

q-fin.PM↗

arXiv2025

Avoidance Loci of Real Projective Varieties

We study real linear spaces in projective space that avoid the real points of a non-degenerate projective variety. For a variety $X \subset \mathbb{P}^{n-1}$ with a real smooth point, we define the avoidance locus $\mathcal{A}_k(X)$ as the subset of the real Grassmannian $\mathrm{Gr}(k,n)_{\mathbb{R}}$ consisting of linear spaces that meet $X$ transversely but contain no real point of $X$. Our construction generalizes the cone of positive polynomials on $\mathbb{R}^n.$ We prove that the avoidance locus is an open semi-algebraic set equal to a union of regions in the complement of a higher Chow form, and that distinct regions are non-adjacent. We present explicit examples for linear spaces, curves, and surfaces, and provide bounds on the number of connected components of $\mathcal{A}_{n-1}(X)$ in terms of the topology of the real locus $X_{\mathbb{R}}$. Finally, we prove that avoidance loci are slice-convex.

math.AG↗

arXiv2025

Scalable Autoregressive Image Generation with Mamba

We introduce AiM, an autoregressive (AR) image generative model based on Mamba architecture. AiM employs Mamba, a novel state-space model characterized by its exceptional performance for long-sequence modeling with linear time complexity, to supplant the commonly utilized Transformers in AR image generation models, aiming to achieve both superior generation quality and enhanced inference speed. Unlike existing methods that adapt Mamba to handle two-dimensional signals via multi-directional scan, AiM directly utilizes the next-token prediction paradigm for autoregressive image generation. This approach circumvents the need for extensive modifications to enable Mamba to learn 2D spatial representations. By implementing straightforward yet strategically targeted modifications for visual generative tasks, we preserve Mamba's core structure, fully exploiting its efficient long-sequence modeling capabilities and scalability. We provide AiM models in various scales, with parameter counts ranging from 148M to 1.3B. On the ImageNet1K 256*256 benchmark, our best AiM model achieves a FID of 2.21, surpassing all existing AR models of comparable parameter counts and demonstrating significant competitiveness against diffusion models, with 2 to 10 times faster inference speed. Code is available at https://github.com/hp-l33/AiM

cs.CV↗