arXiv subjects

Wang Yu

Publications and source records attributed to Wang Yu.

8 recordsLinked to original sources

arXiv2026

Adapting Technical-Service LLM Agents with Latent Logic Augmentation, Robust Noise Reduction, and Hybrid Reward Modeling

Technical-service LLM agents are entering production workflows, where value depends on whether engineers adopt generated replies. Service tickets hide decision logic, contain noisy single-reference responses, and make reward evaluation costly, making standard post-training brittle. Existing post-training and LLM-as-a-Judge approaches improve grounding or feedback, but do not jointly model latent decision logic, response diversity, and reward cost. We address this gap by coupling latent logic augmentation, robust noise reduction, and hybrid reward modeling. The framework augments supervised fine-tuning data with Planning-Aware Trajectory Modeling and Reasoning Augmentation, builds dual-filtered Multiple Ground Truths, and trains the policy with a hybrid reward that combines a Reranker with an LLM-as-a-Judge. On real Cloud technical-service tasks, the adapted Qwen3-4B achieves the highest Multi-ECS (0.441), lower reward cost, and the highest production adoption rate (46.63%).

cs.LG↗

arXiv2026

DynHD: Hallucination Detection for Diffusion Large Language Models via Denoising Dynamics Deviation Learning

Diffusion large language models (D-LLMs) have emerged as a promising alternative to auto-regressive models due to their iterative refinement capabilities. However, hallucinations remain a critical issue that hinders their reliability. To detect hallucination responses from model outputs, token-level uncertainty, such as entropy, has been widely used to indicate potential factual errors. Nevertheless, unlike auto-regressive models that generate tokens sequentially, D-LLMs generate fixed-length sequences simultaneously, where only a small subset of tokens is informative for hallucination detection. Thus, aggregating uncertainty over all tokens can be suboptimal. Moreover, the evolution trend of uncertainty throughout the diffusion process can also provide valuable signals, highlighting the necessity of modeling its denoising dynamics for hallucination detection. In this paper, we propose DynHD, which bridges these gaps from both spatial (token sequence) and temporal (denoising dynamics) perspectives. To handle the information density imbalance across tokens, we propose a semantic-aware evidence construction module that extracts hallucination-indicative signals by removing task-invariant structural tokens and emphasizing the uncertainty of the remaining informative tokens. To model denoising dynamics for hallucination detection, we introduce a reference evidence generator that learns the expected evolution trajectory of uncertainty evidence, along with a deviation-based hallucination detector that makes predictions by measuring the discrepancy between the observed and reference trajectories. Extensive experiments demonstrate that DynHD consistently outperforms state-of-the-art baselines while achieving higher efficiency across multiple benchmarks and backbone models. The code is available at: https://github.com/qyy11-com/DynHD.

cs.CL↗

arXiv2024

Local and Global Feature Attention Fusion Network for Face Recognition

Recognition of low-quality face images remains a challenge due to invisible or deformation in partial facial regions. For low-quality images dominated by missing partial facial regions, local region similarity contributes more to face recognition (FR). Conversely, in cases dominated by local face deformation, excessive attention to local regions may lead to misjudgments, while global features exhibit better robustness. However, most of the existing FR methods neglect the bias in feature quality of low-quality images introduced by different factors. To address this issue, we propose a Local and Global Feature Attention Fusion (LGAF) network based on feature quality. The network adaptively allocates attention between local and global features according to feature quality and obtains more discriminative and high-quality face features through local and global information complementarity. In addition, to effectively obtain fine-grained information at various scales and increase the separability of facial features in high-dimensional space, we introduce a Multi-Head Multi-Scale Local Feature Extraction (MHMS) module. Experimental results demonstrate that the LGAF achieves the best average performance on $4$ validation sets (CFP-FP, CPLFW, AgeDB, and CALFW), and the performance on TinyFace and SCFace outperforms the state-of-the-art methods (SoTA).

cs.CV↗

arXiv2023

An Efficient Quantum Circuit Construction Method for Mutually Unbiased Bases in $n$-Qubit Systems

Mutually unbiased bases (MUBs) play a crucial role in numerous applications within quantum information science, such as quantum state tomography, error correction, entanglement detection, and quantum cryptography. Utilizing $2^n + 1$ MUB circuits provides a minimal and optimal measurement strategy for reconstructing all $n$-qubit unknown states. It significantly reduces the number of measurements compared to the traditional $4^n$ Pauli observables, also enhancing the robustness of quantum key distribution (QKD) protocols. Previous circuit designs that rely on a single generator can result in exponential gate costs for some MUB circuits. In this work, we present an efficient algorithm to generate each of the $2^n + 1$ quantum MUB circuits on $n$-qubit systems within $O(n^3)$ time. The algorithm features a three-stage structure, and we have calculated the average number of different gates for random sampling. Additionally, we have identified two linear properties: the entanglement part can be directly defined into $2n - 3$ fixed sub-parts, and the knowledge of $n$ special MUB circuits is sufficient to construct all $2^n + 1$ MUB circuits. This new efficient and simple circuit construction paves the way for the implementation of a complete set of MUBs in diverse quantum information processing tasks on high-dimensional quantum systems.

quant-ph↗

arXiv2021

CIIA:A New Algorithm for Community Detection

In this paper, through thinking on the modularity function that measures the standard of community division, a new algorithm for dividing communities is proposed, called the Connect Intensity Iteration algorithm, or CIIA for short. In this algorithm, a new indicator is proposed.This indicator is the difference between the actual number of edges between two nodes and the number of edges when the edges are randomly placed. It can reflect more information between the nodes. The larger the value of this index, the greater the possibility that the two nodes are divided into the same community, and vice versa. This paper also verifies the algorithm through numerical simulations and real cases, and the results show the feasibility of the algorithm.

cs.SI↗

arXiv2021

ELM-based Frame Synchronization in Nonlinear Distortion Scenario Using Superimposed Training

The requirement of high spectrum efficiency puts forward higher requirements on frame synchronization (FS) in wireless communication systems. Meanwhile, a large number of nonlinear devices or blocks will inevitably cause nonlinear distortion. To avoid the occupation of bandwidth resources and overcome the difficulty of nonlinear distortion, an extreme learning machine (ELM)-based network is introduced into the superimposed training-based FS with nonlinear distortion. Firstly, a preprocessing procedure is utilized to reap the features of synchronization metric (SM). Then, based on the rough features of SM, an ELM network is constructed to estimate the offset of frame boundary. The analysis and experiment results show that, compared with existing methods, the proposed method can improve the error probability of FS and bit error rate (BER) of symbol detection (SD). In addition, this improvement has its robustness against the impacts of parameter variations.

eess.SP↗

arXiv2020

ELM-based Frame Synchronization in Burst-Mode Communication Systems with Nonlinear Distortion

In burst-mode communication systems, the quality of frame synchronization (FS) at receivers significantly impacts the overall system performance. To guarantee FS, an extreme learning machine (ELM)-based synchronization method is proposed to overcome the nonlinear distortion caused by nonlinear devices or blocks. In the proposed method, a preprocessing is first performed to capture the coarse features of synchronization metric (SM) by using empirical knowledge. Then, an ELM-based FS network is employed to reduce system's nonlinear distortion and improve SMs. Experimental results indicate that, compared with existing methods, our approach could significantly reduce the error probability of FS while improve the performance in terms of robustness and generalization.

eess.SP↗

arXiv2014

A linear calibration method on DNL error for energy spectrum

A calibration method aimed for the differential nonlinearity (DNL) error of the Low Energy X-ray Instrument (LE) onboard the Hard X-ray Modulation Telescope (HXMT) is presented, which is independent with electronic systems used as testing platform and is only determined by the analog-to-digital converter (ADC) itself. Exploring this method, ADCs that are used within the flight model phase of HXMT-LE can be calibrated individually and independently by a non-destructive and low-cost way, greatly alleviating the complexity of the problem. As a result, the performance of the energy spectrum can be significantly improved, further more, noise reduced and resolution enhanced.

astro-ph.IM↗