arXiv subjects

Yufeng Zhou

Publications and source records attributed to Yufeng Zhou.

12 recordsLinked to original sources

arXiv2026

Topological Quantum Interferometry

Structured light provides high-dimensional Hilbert spaces holding tremendous potential for fundamental quantum optics and quantum technologies. However, existing characterization methods, like Hong-Ou-Mandel (HOM) interference, typically assume perfectly tuned conditions, overlooking the geometric physics governing spatial mode evolution. Here, we establish topological quantum interferometry driven by an interaction-based geometric phase, the exchange Berry phase (BPX). Our formalism generalizes $q$-plate state generation and characterization to arbitrary topological charges and (de)tuning conditions, demonstrating that BPX acts as a geometric marker governing spatial interference. We show BPX serves as a deterministic control parameter, decomposing two-photon spatial patterns into geometry-dictated fundamental modes. This mapping reveals topological invariants and phase singularities that function as a non-tomographic witness for state dimensionality estimation, circumventing full-state reconstruction. Being device-independent and highly scalable, this approach enables scalable high-dimensional characterization and topologically protected state selection, with direct applicability to quantum metrology and high-capacity quantum networks.

quant-ph↗

arXiv2026

SOLAR-RL: Semi-Online Long-horizon Assignment Reinforcement Learning

As Multimodal Large Language Models (MLLMs) mature, GUI agents are evolving from static interactions to complex navigation. While Reinforcement Learning (RL) has emerged as a promising paradigm for training MLLM agents on dynamic GUI tasks, its effective application faces a dilemma. Standard Offline RL often relies on static step-level data, neglecting global trajectory semantics such as task completion and execution quality. Conversely, Online RL captures the long-term dynamics but suffers from high interaction costs and potential environmental instability. To bridge this gap, we propose SOLAR-RL (Semi-Online Long-horizon Assignment Reinforcement Learning). Instead of relying solely on expensive online interactions, our framework integrates global trajectory insights directly into the offline learning process. Specifically, we reconstruct diverse rollout candidates from static data, detect the first failure point using per-step validity signals, and retroactively assign dense step-level rewards with target-aligned shaping to reflect trajectory-level execution quality, effectively simulating online feedback without interaction costs. Extensive experiments demonstrate that SOLAR-RL significantly improves long-horizon task completion rates and robustness compared to strong baselines, offering a sample-efficient solution for autonomous GUI navigation.

cs.LG↗

arXiv2026

Skill-SD: Skill-Conditioned Self-Distillation for Multi-turn LLM Agents

Reinforcement learning (RL) has been widely used to train LLM agents for multi-turn interactive tasks, but its sample efficiency is severely limited by sparse rewards and long horizons. On-policy self-distillation (OPSD) alleviates this by providing dense token-level supervision from a privileged teacher that has access to ground-truth answers. However, such fixed privileged information cannot capture the diverse valid strategies in agent tasks, and naively combining OPSD with RL often leads to training collapse. To address these limitations, we introduce Skill-SD, a framework that turns the agent's own trajectories into dynamic training-only supervision. Completed trajectories are summarized into compact natural language skills that describe successful behaviors, mistakes, and workflows. These skills serve as dynamic privileged information conditioning only the teacher, while the student always acts under the plain task prompt and learns to internalize the guidance through distillation. To stabilize the training, we derive an importance-weighted reverse-KL loss to provide gradient-correct token-level distillation, and dynamically synchronize the teacher with the improving student. Experimental results on agentic benchmarks demonstrate that Skill-SD substantially outperforms the standard RL baseline, improving both vanilla GRPO (+14.0%/+10.9% on AppWorld/Sokoban) and vanilla OPD (+42.1%/+40.6%). Project page: https://k1xe.github.io/skill-sd/

cs.LG↗

arXiv2026

On the Computation Rate of All-Reduce

In the All-Reduce problem, each one of the K nodes holds an input and wishes to compute the sum of all K inputs through a communication network where each pair of nodes is connected by a parallel link with arbitrary bandwidth. The computation rate of All-Reduce is defined as the number of sum instances that can be computed over each network use. For the computation rate, we provide a cut-set upper bound and a linear programming lower bound based on time (bandwidth) sharing over all schemes that first perform Reduce (aggregating all inputs at one node) and then perform Broadcast (sending the sum from that node to all other nodes). Specializing the two general bounds gives us the optimal computation rate for a class of communication networks and the best-known rate bounds (where the upper bound is no more than twice of the lower bound) for cyclic, complete, and hypercube networks.

cs.IT↗

arXiv2026

UI-Mem: Self-Evolving Experience Memory for Online Reinforcement Learning in Mobile GUI Agents

Online Reinforcement Learning (RL) offers a promising paradigm for enhancing GUI agents through direct environment interaction. However, its effectiveness is severely hindered by inefficient credit assignment in long-horizon tasks and repetitive errors across tasks due to the lack of experience transfer. To address these challenges, we propose UI-Mem, a novel framework that enhances GUI online RL with a Hierarchical Experience Memory. Unlike traditional replay buffers, our memory accumulates structured knowledge, including high-level workflows, subtask skills, and failure patterns. These experiences are stored as parameterized templates that enable cross-task and cross-application transfer. To effectively integrate memory guidance into online RL, we introduce Stratified Group Sampling, which injects varying levels of guidance across trajectories within each rollout group to maintain outcome diversity, driving the unguided policy toward internalizing guided behaviors. Furthermore, a Self-Evolving Loop continuously abstracts novel strategies and errors to keep the memory aligned with the agent's evolving policy. Experiments on online GUI benchmarks demonstrate that UI-Mem significantly outperforms traditional RL baselines and static reuse strategies, with strong generalization to unseen applications. Project page: https://ui-mem.github.io

cs.CV↗

arXiv2026

The Dynamic Team Orienteering Problem in Spatial Crowdsourcing: A Scenario Sampling Approach

In services such as retail audits and urban infrastructure monitoring, a platform dispatches rewarded, location-based micro-tasks to mobile workers traveling along personal origin-destination (OD) trips under hard time budgets. As requests with time constraints arrive online over a finite horizon, the platform must decide which requests to accept and how to route workers to maximize collected profit. We model this setting as the Dynamic Team Orienteering Problem in Spatial Crowdsourcing (DTOP-SC). To solve this problem, we propose a scenario-sampling rolling-horizon framework that mitigates myopic bias by augmenting each planning epoch with sampled virtual tasks. At each epoch, the augmented task set defines a deterministic static subproblem solved via an adaptive large neighborhood search (ALNS). We also formulate a mixed-integer programming model to provide offline reference solutions. Computational experiments are conducted on synthetic DTOP-SC instances generated from real-world road-map coordinates and on a dynamic team orienteering (DTOP) benchmark. On the map-based instances, the proposed policy exhibits stable gaps with respect to time-limited MIP solutions across the tested scales, while maintaining smooth computational scalability as the problem size increases. On the DTOP benchmark, the policy achieves an average decision time of 0.14s per instance, with 192-198s reported for multiple plan approach as an indicative reference, while maintaining competitive profit.

math.OC↗

arXiv2025

A survey of the orienteering problem: model evolution, algorithmic advances, and future directions

The orienteering problem (OP) is a combinatorial optimization problem that seeks a path visiting a subset of locations to maximize collected rewards under a limited resource budget. This article presents a systematic PRISMA-based review of OP research published between 2017 and 2025, with a focus on models and methods that have shaped subsequent developments in the field. We introduce a component-based taxonomy that decomposes OP variants into time-, path-, node-, structure-, and information-based extensions. This framework unifies classical and emerging variants -- including stochastic, time-dependent, Dubins, Set, and multi-period OPs -- within a single structural perspective. We further categorize solution approaches into exact algorithms, heuristics and metaheuristics, and learning-based methods, with particular emphasis on matheuristics and recent advances in artificial intelligence, especially reinforcement learning and neural networks, which enhance scalability in large-scale and information-rich settings. Building on this unified view, we discuss how different components affect computational complexity and polyhedral properties and identify open challenges related to robustness, sustainability, and AI integration. The survey thus provides both a consolidated reference for existing OP research and a structured agenda for future theoretical and applied work.

math.OC↗

arXiv2025

Joint Optimization for Multi-User Transmissive RIS-MIMO Systems

Transmissive reconfigurable intelligent surfaces (RIS) represent a transformative architecture for future wireless networks, enabling a paradigm shift from traditional costly base stations to low-cost, energy-efficient transmitters. This paper explores a downlink multi-user MIMO system where a transmissive RIS, illuminated by a single feed antenna, forms the core of the transmitter. The joint optimization of the RIS coefficient vector, power allocation, and receive beamforming in such a system is critical for performance but poses significant challenges due to the non-convex objective, coupled variables, and constant modulus constraints. To address these challenges, we propose a novel optimization framework. Our approach involves reformulating the sum-rate maximization problem into a tractable equivalent form and developing an efficient alternating optimization (AO) algorithm. This algorithm decomposes the problem into subproblems for the RIS coefficients, receive beamformers, and power allocation, each solved using advanced techniques including convex approximation and difference-of-convex programming. Simulation results demonstrate that our proposed method converges rapidly and achieves substantial sum-rate gains over conventional schemes, validating the effectiveness of our approach and highlighting the potential of transmissive RIS as a key technology for next-generation wireless systems.

cs.IT↗

arXiv2025

Exploiting Movable Antennas in NOMA Networks: Joint Beamforming, Power Allocation and Antenna Position Optimization

This paper investigates the movable antenna (MA)- assisted downlink non-orthogonal multiple access (NOMA) network to maximize system throughput. In the considered scenario, both the base station (BS) and users are equipped with MA, and a predetermined successive interference cancellation (SIC) decoding order is adopted. Based on the field-response channel model, we formulate a complex, non-convex problem to jointly optimize the BS beamforming, power allocation, and MA positions at both the transmitter and receivers. To address this, we propose an efficient algorithm based on an alternating optimization (AO) framework, which decomposes the original problem into three distinct subproblems. By employing sequential parametric convex approximation (SPCA) and successive convex approximation (SCA) techniques, the non-convex constraints within each subproblem are transformed into tractable. This methodology ensures the algorithm converges to a stable, locally optimal solution. Numerical results validate that the proposed system, which fully exploits the degrees of freedom from antenna mobility at both ends, significantly outperforms benchmarks in terms of throughput.

eess.SP↗

arXiv2024

H-SGANet: Hybrid Sparse Graph Attention Network for Deformable Medical Image Registration

The integration of Convolutional Neural Network (ConvNet) and Transformer has emerged as a strong candidate for image registration, leveraging the strengths of both models and a large parameter space. However, this hybrid model, treating brain MRI volumes as grid or sequence structures, faces challenges in accurately representing anatomical connectivity, diverse brain regions, and vital connections contributing to the brain's internal architecture. Concerns also arise regarding the computational expense and GPU memory usage associated with this model. To tackle these issues, a lightweight hybrid sparse graph attention network (H-SGANet) has been developed. This network incorporates a central mechanism, Sparse Graph Attention (SGA), based on a Vision Graph Neural Network (ViG) with predetermined anatomical connections. The SGA module expands the model's receptive field and seamlessly integrates into the network. To further amplify the advantages of the hybrid network, the Separable Self-Attention (SSA) is employed as an enhanced token mixer, integrated with depth-wise convolution to constitute SSAFormer. This strategic integration is designed to more effectively extract long-range dependencies. As a hybrid ConvNet-ViG-Transformer model, H-SGANet offers threefold benefits for volumetric medical image registration. It optimizes fixed and moving images concurrently through a hybrid feature fusion layer and an end-to-end learning framework. Compared to VoxelMorph, a model with a similar parameter count, H-SGANet demonstrates significant performance enhancements of 3.5% and 1.5% in Dice score on the OASIS dataset and LPBA40 dataset, respectively.

cs.CV↗

arXiv2024

Movable Antenna Empowered Downlink NOMA Systems: Power Allocation and Antenna Position Optimization

This paper investigates a novel communication paradigm employing movable antennas (MAs) within a multiple-input single-output (MISO) non-orthogonal multiple access (NOMA) downlink framework, where users are equipped with MAs. Initially, leveraging the far-field response, we delineate the channel characteristics concerning both the power allocation coefficient and positions of MAs. Subsequently, we endeavor to maximize the channel capacity by jointly optimizing power allocation and antenna positions. To tackle the resultant non-convex problem, we propose an alternating optimization (AO) scheme underpinned by successive convex approximation (SCA) to converge towards a stationary point. Through numerical simulations, our findings substantiate the superiority of the MA-assisted NOMA system over both orthogonal multiple access (OMA) and conventional NOMA configurations in terms of average sum rate and outage probability.

cs.IT↗

arXiv2022

Power Efficient Video Super-Resolution on Mobile NPUs with Deep Learning, Mobile AI & AIM 2022 challenge: Report

Video super-resolution is one of the most popular tasks on mobile devices, being widely used for an automatic improvement of low-bitrate and low-resolution video streams. While numerous solutions have been proposed for this problem, they are usually quite computationally demanding, demonstrating low FPS rates and power efficiency on mobile devices. In this Mobile AI challenge, we address this problem and propose the participants to design an end-to-end real-time video super-resolution solution for mobile NPUs optimized for low energy consumption. The participants were provided with the REDS training dataset containing video sequences for a 4X video upscaling task. The runtime and power efficiency of all models was evaluated on the powerful MediaTek Dimensity 9000 platform with a dedicated AI processing unit capable of accelerating floating-point and quantized neural networks. All proposed solutions are fully compatible with the above NPU, demonstrating an up to 500 FPS rate and 0.2 [Watt / 30 FPS] power consumption. A detailed description of all models developed in the challenge is provided in this paper.

eess.IV↗