arXiv subjects

Yuying Liu

Publications and source records attributed to Yuying Liu.

14 recordsLinked to original sources

arXiv2026

Magnetic Breakdown Reshapes Quantum Oscillations in Kagome Metals

Recent quantum-oscillation experiments on kagome metals have revealed markedly different phase offsets even among systems with nearly identical band structures and Fermi-surface geometries. Using a tight-binding model, we show that weak orbital hybridization can slightly modify the hybridization gaps. Small variations in these gaps can substantially alter the measured oscillation phase, despite leaving the overall electronic structure nearly unchanged. This phase shift originates from magnetic breakdown, which reconstructs cyclotron trajectories and can mask the nontrivial phase of an isolated orbit, yielding a trivial phase offset. Moreover, uniaxial strain can tune the relevant hybridization gaps and thereby weaken magnetic breakdown. This restores the nontrivial phase offset that magnetic breakdown otherwise masks, providing an experimentally accessible knob for controlling the oscillation phase. These results identify magnetic breakdown as the key mechanism controlling the phase shift and provide a plausible explanation for recent experimental phase discrepancies in kagome metals.

cond-mat.mes-hall↗

arXiv2025

MMRAG-RFT: Two-stage Reinforcement Fine-tuning for Explainable Multi-modal Retrieval-augmented Generation

Multi-modal Retrieval-Augmented Generation (MMRAG) enables highly credible generation by integrating external multi-modal knowledge, thus demonstrating impressive performance in complex multi-modal scenarios. However, existing MMRAG methods fail to clarify the reasoning logic behind retrieval and response generation, which limits the explainability of the results. To address this gap, we propose to introduce reinforcement learning into multi-modal retrieval-augmented generation, enhancing the reasoning capabilities of multi-modal large language models through a two-stage reinforcement fine-tuning framework to achieve explainable multi-modal retrieval-augmented generation. Specifically, in the first stage, rule-based reinforcement fine-tuning is employed to perform coarse-grained point-wise ranking of multi-modal documents, effectively filtering out those that are significantly irrelevant. In the second stage, reasoning-based reinforcement fine-tuning is utilized to jointly optimize fine-grained list-wise ranking and answer generation, guiding multi-modal large language models to output explainable reasoning logic in the MMRAG process. Our method achieves state-of-the-art results on WebQA and MultimodalQA, two benchmark datasets for multi-modal retrieval-augmented generation, and its effectiveness is validated through comprehensive ablation experiments.

cs.AI↗

arXiv2025

SegQuant: A Semantics-Aware and Generalizable Quantization Framework for Diffusion Models

Diffusion models have demonstrated exceptional generative capabilities but are computationally intensive, posing significant challenges for deployment in resource-constrained or latency-sensitive environments. Quantization offers an effective means to reduce model size and computational cost, with post-training quantization (PTQ) being particularly appealing due to its compatibility with pre-trained models without requiring retraining or training data. However, existing PTQ methods for diffusion models often rely on architecture-specific heuristics that limit their generalizability and hinder integration with industrial deployment pipelines. To address these limitations, we propose SegQuant, a unified quantization framework that adaptively combines complementary techniques to enhance cross-model versatility. SegQuant consists of a segment-aware, graph-based quantization strategy (SegLinear) that captures structural semantics and spatial heterogeneity, along with a dual-scale quantization scheme (DualScale) that preserves polarity-asymmetric activations, which is crucial for maintaining visual fidelity in generated outputs. SegQuant is broadly applicable beyond Transformer-based diffusion models, achieving strong performance while ensuring seamless compatibility with mainstream deployment tools.

cs.CV↗

arXiv2025

RLJP: Legal Judgment Prediction via First-Order Logic Rule-enhanced with Large Language Models

Legal Judgment Prediction (LJP) is a pivotal task in legal AI. Existing semantic-enhanced LJP models integrate judicial precedents and legal knowledge for high performance. But they neglect legal reasoning logic, a critical component of legal judgments requiring rigorous logical analysis. Although some approaches utilize legal reasoning logic for high-quality predictions, their logic rigidity hinders adaptation to case-specific logical frameworks, particularly in complex cases that are lengthy and detailed. This paper proposes a rule-enhanced legal judgment prediction framework based on first-order logic (FOL) formalism and comparative learning (CL) to develop an adaptive adjustment mechanism for legal judgment logic and further enhance performance in LJP. Inspired by the process of human exam preparation, our method follows a three-stage approach: first, we initialize judgment rules using the FOL formalism to capture complex reasoning logic accurately; next, we propose a Confusion-aware Contrastive Learning (CACL) to dynamically optimize the judgment rules through a quiz consisting of confusable cases; finally, we utilize the optimized judgment rules to predict legal judgments. Experimental results on two public datasets show superior performance across all metrics. The code is publicly available{https://anonymous.4open.science/r/RLJP-FDF1}.

cs.AI↗

arXiv2025

On bifurcation of a stage-structured single-species model with harvest

This paper investigates the dynamics of the Nicholson's blowffies equation with stage structure and harvest. By employing the property of Lambert W function, the existence of positive equilibria is obtained. With aid of the distribution of the eigenvalues in the characteristic equation, the local stability of the equilibria and the existence of Hopf bifurcation of the singlespecies model are obtained. Furthermore, by applying the results due to Balazs I., Rost G. (Internat. J. Bifur. Chaos 31(2021):2150071), when the harvest rate is sufffciently small, the direction of the Hopf bifurcations at the ffrst and last bifurcation values are forward and backward, respectively, and the bifurcating periodic solutions are all asymptotically stable. Finally, Numerical simulations are conducted to validate the theoretical conclusions. These results can be seen as the complement of the works of Shu et al. (J. Differential Equations 255 (2013) 2565).

math.DS↗

arXiv2022

Physics-Informed Koopman Network

Koopman operator theory is receiving increased attention due to its promise to linearize nonlinear dynamics. Neural networks that are developed to represent Koopman operators have shown great success thanks to their ability to approximate arbitrarily complex functions. However, despite their great potential, they typically require large training data-sets either from measurements of a real system or from high-fidelity simulations. In this work, we propose a novel architecture inspired by physics-informed neural networks, which leverage automatic differentiation to impose the underlying physical laws via soft penalty constraints during model training. We demonstrate that it not only reduces the need of large training data-sets, but also maintains high effectiveness in approximating Koopman eigenfunctions.

cs.LG↗

arXiv2020

A relaxation accelerated two-sweep modulus-based matrix splitting iteration method for solving linear complementarity problems

For a linear complementarity problem, we present a relaxaiton accelerated two-sweep matrix splitting iteration method. The convergence analysis illustrates that the proposed method converges to the exact solution of the linear complementarity problem when the system matrix is an $H_+$-matrix and the convergence conditions are given. Numerical experiments show that the proposed method is more efficient than the existing ones.

math.OC↗

arXiv2020

Hierarchical Deep Learning of Multiscale Differential Equation Time-Steppers

Nonlinear differential equations rarely admit closed-form solutions, thus requiring numerical time-stepping algorithms to approximate solutions. Further, many systems characterized by multiscale physics exhibit dynamics over a vast range of timescales, making numerical integration computationally expensive due to numerical stiffness. In this work, we develop a hierarchy of deep neural network time-steppers to approximate the flow map of the dynamical system over a disparate range of time-scales. The resulting model is purely data-driven and leverages features of the multiscale dynamics, enabling numerical integration and forecasting that is both accurate and highly efficient. Moreover, similar ideas can be used to couple neural network-based models with classical numerical time-steppers. Our multiscale hierarchical time-stepping scheme provides important advantages over current time-stepping algorithms, including (i) circumventing numerical stiffness due to disparate time-scales, (ii) improved accuracy in comparison with leading neural-network architectures, (iii) efficiency in long-time simulation/forecasting due to explicit training of slow time-scale dynamics, and (iv) a flexible framework that is parallelizable and may be integrated with standard numerical time-stepping algorithms. The method is demonstrated on a wide range of nonlinear dynamical systems, including the Van der Pol oscillator, the Lorenz system, the Kuramoto-Sivashinsky equation, and fluid flow pass a cylinder; audio and video signals are also explored. On the sequence generation examples, we benchmark our algorithm against state-of-the-art methods, such as LSTM, reservoir computing, and clockwork RNN. Despite the structural simplicity of our method, it outperforms competing methods on numerical integration.

cs.LG↗

arXiv2020

Multiresolution Convolutional Autoencoders

We propose a multi-resolution convolutional autoencoder (MrCAE) architecture that integrates and leverages three highly successful mathematical architectures: (i) multigrid methods, (ii) convolutional autoencoders and (iii) transfer learning. The method provides an adaptive, hierarchical architecture that capitalizes on a progressive training approach for multiscale spatio-temporal data. This framework allows for inputs across multiple scales: starting from a compact (small number of weights) network architecture and low-resolution data, our network progressively deepens and widens itself in a principled manner to encode new information in the higher resolution data based on its current performance of reconstruction. Basic transfer learning techniques are applied to ensure information learned from previous training steps can be rapidly transferred to the larger network. As a result, the network can dynamically capture different scaled features at different depths of the network. The performance gains of this adaptive multiscale architecture are illustrated through a sequence of numerical experiments on synthetic examples and real-world spatial-temporal data.

cs.LG↗

arXiv2017

Exponential Change of Measure for General Piecewise Deterministic Markov Processes

We consider a general piecewise deterministic Markov process (PDMP) $X=\{X_t\}_{t\geqslant 0}$ with measure-valued generator $\mathcal{A}$, for which the conditional distribution function of the inter-occurrence time is not necessarily absolutely continuous. A general form of the exponential martingales is presented as $$M^f_t=\frac{f(X_t)}{f(X_0)}\left[\mathrm{Sexp}\left(\int_{(0,t]}\frac{\mathrm{d}L(\mathcal{A}f)_s}{f(X_{s-})}\right)\right]^{-1}.$$ Using this exponential martingale as a likelihood ratio process, we define a new probability measure. It is shown that the original process remains a general PDMP under the new probability measure. And we find the new measure-valued generator and its domain.

math.PR↗

arXiv2015

Study of Calibration of Solar Radio Spectrometers and the quiet-Sun Radio Emission

This work presents a systematic investigation of the influence of weather conditions on the calibration errors by using Gaussian fitness, least chi-square linear fitness and wavelet transform to analyze the calibration coefficients from observations of the Chinese Solar Broadband Radio Spectrometers (at frequency bands of 1.0-2.0 GHz, 2.6-3.8 GHz, and 5.2-7.6 GHz) during 1997-2007. We found that calibration coefficients are influenced by the local air temperature. Considering the temperature correction, the calibration error will reduce by about $10\%-20\%$ at 2800 MHz. Based on the above investigation and the calibration corrections, we further study the radio emission of the quiet-Sun by using an appropriate hybrid model of the quiet-Sun atmosphere. The results indicate that the numerical flux of the hybrid model is much closer to the observation flux than that of other ones.

astro-ph.SR↗

arXiv2012

Microwave burst with fine spectral structures in a solar flare on 2011 August 9

On August 9, 2011, there was an X6.9 flare event occurred near the west limb of solar disk. From the observation obtained by the spectrometer of the Chinese Solar Broadband Radio Spectrometer in Huairou (SBRS/Huairou) around the flare, we find that this powerful flare has only a short-duration microwave burst of about only 5 minutes, and during the short-duration microwave burst, there are several kinds of fine structures on the spectrogram. These fine structures include very short-period pulsations, millisecond spike bursts, and type III bursts. The most interesting is that almost all of the pulses of very short-period pulsation (VSP) are structured by clusters of millisecond timescales of spike bursts or type III bursts. And there exists three different kinds of frequency drift rates in the VSP: the frequency drift rates with absolute value of about 55 - 130 MHz s^{-1} in the pulse groups, the frequency drift rates with absolute value of about 2.91 - 16.9 GHz s^{-1} on each individual pulse, and the frequency drift rates with absolute value of about 15 - 25 GHz s$^{-1}$) at each individual spike burst or type III burst.

astro-ph.SR↗

arXiv2010

Microwave Quasi-periodic Pulsations in Multi-timescales Associated with a Solar Flare/CME Event

Microwave observations of quasi-periodic pulsations (QPP) in multi-timescales are confirmed to be associated with an X3.4 flare/CME event at Solar Broadband Radio Spectrometer in Huairou (SBRS/Huairou) on 13 December 2006. It is most remarkable that the timescales of QPPs are distributed in a broad range from hecto-second (very long period pulsation, VLP, the period P>100 s), deca-second (long period pulsation, LPP, 10<P<100 s), few seconds (short period pulsation, SPP, 1<P<10 s), deci-second (slow-very short period pulsation, slow-VSP, 0.1<P<1.0 s), to centi-second (fast-very short period pulsation, fast-VSP, P<0.1 s), and forms a broad hierarchy of timescales. The statistical distribution in logarithmic period-duration space indicates that QPPs can be classified into two groups: group I includes VLP, LPP, SPP and part of slow-VSPs distributed around a line approximately; group II includes fast-VSP and most of slow-VSP dispersively distributed away from the above line. This feature implies that the generation mechanism of group I is different from group II. Group I is possibly related with some MHD oscillations in magnetized plasma loops in the active region, e.g., VLP may be generated by standing slow sausage mode coupling and resonating with the underlying photospheric 5-min oscillation, the modulation is amplified and forms the main framework of the whole flare/CME process; LPP, SPP, and part of slow-VSPs are most likely to be caused by standing fast modes or LRC-circuit resonance in current-carrying plasma loops. Group II is possibly generated by modulations of resistive tearing-mode oscillations in electric current-carrying flaring loops.

astro-ph.SR↗

arXiv2009

Broadband Radio Spectral Observations of Solar Eclipse on 2008-08-01 and Implications on the Quiet Sun Atmospheric Model

Based on the joint-observations of the radio broadband spectral emissions of solar eclipse on August 1, 2008 at Jiuquan (total eclipse) and Huairou (partial eclipse) at the frequencies of 2.00 -- 5.60 GHz (Jiuquan), 2.60 -- 3.80 GHZ (Chinese solar broadband radiospectrometer, SBRS/Huairou), and 5.20 -- 7.60 GHz (SBRS/Huairou), the authors assemble a successive series of broadband spectrum with a frequency of 2.60 -- 7.60 GHz to observe the solar eclipse synchronously. This is the first attempt to analyze the solar eclipse radio emission under the two telescopes located at different places with broadband frequencies in the periods of total and partial eclipse. With these analyses, the authors made a new semiempirical model of the coronal plasma density of the quiet Sun and made a comparison with the classic models.

astro-ph.SR↗