Searcharxiv⌕ Search

arXiv subjects

Ping Li

Publications and source records attributed to Ping Li.

At least 37 records · Page 2Linked to original sources
arXiv2026

Applications of the Transformer Architecture in AI-Assisted English Reading Comprehension

This paper studies interpretable and fair artificial intelligence architectures for understanding English reading. Introduced transformer-based models, integrating advanced attention mechanisms and gradient-based feature attribution. The model's lack of interpretability, reduction of algorithmic bias, and unreliable performance in learning environments are the current issues faced in natural language teaching. A unified technical pipeline has been constructed, including adversarial bias correction methods, token-level attribution analysis, and multi-head attention heatmap visualization. Experimental validation was conducted using a large-scale labeled English reading comprehension dataset, and the data partitioning scheme and parameter optimization procedures have been determined. The method significantly outperforms the state-of-the-art models for this task in terms of accuracy and macro-average F1 score; in some aspects, it even surpasses or closely matches the results of human evaluations. In multi-week user experiments, the explainable transformer improved teachers' trust and operability in feedback-based assessments within the scoring system. The proposed method aims to ensure high prediction accuracy and fairness for different learners. This indicates that it is a real-world educational application based on artificial intelligence with a focus on interpretation. Improve the user experience in AI-assisted reading comprehension systems, counteract biases, and enhance the details explained by transformers.

cs.CL↗
arXiv2026

Pumping of spin supercurrent in unitary triplet superconductors

One efficient mechanism for generating a charge supercurrent is Andreev reflection, in which the electric current injected from a normal metal into a conventional superconductor is converted into a supercurrent, thereby preserving charge conservation. We here propose a general principle for generating spin supercurrents in triplet superconductors by analogy with such charge transport, i.e., assuming spin conservation. We find a spin torque that is proportional to the triplet superconducting order parameter and, in the spin-conservation scenario, converts the particle spin to that of Cooper pairs. Based on this general principle, we propose an implementation to efficiently generate a spin supercurrent in unitary triplet superconductors, even though Cooper pairs carry no spin polarization at equilibrium, by the magnetization dynamics ${\bf M}(t)$ of a proximity magnetic nanostructure. The efficiency of this spin pumping is not solely limited to the $d{\bf M}/dt\times {\bf M}$ due to the emergent particle-hole symmetry, thereby going beyond the conventional spin pumping of electrons. This general principle provides an efficient approach to generating and manipulating dissipationless spin currents in many unconventional superconductors.

cond-mat.supr-con↗
arXiv2026

The $χ_y$-genus, Chern number inequalities and signature

This article has two parts. In the first part we introduce two positivity conditions for the modified $χ_y$-genus on almost-complex manifolds and show that each of them implies a family of optimal Chern number inequalities. It turns out that many important Kähler and symplectic manifolds satisfy either of the two positivity conditions, and hence these Chern number inequalities hold true on them. In the second part we focus on the signature, a special value of the $χ_y$-genus, of symplectic manifolds equipped with symplectic circle actions and give applications. Our results in this part unify and generalize various related results in the existing literature.

math.DG↗
arXiv2026

SZ Lyncis: A Non-Accreting Neutron Star-delta Scuti Binary Candidate Discovered via Dynamics and Asteroseismology

Neutron stars (NSs) are traditionally discovered through radio, X-ray, or gamma-ray observations, but optical time-domain surveys can unveil non-accreting NSs in wide binaries. Here we report a NS candidate in the single-lined binary SZ~Lyncis, identified through a combination of asteroseismology, spectroscopy, pulsation timing, and astrometry. The visible $δ$ Scuti primary has a mass of $M_1 = 1.83_{-0.01}^{+0.06}~\mathrm{M_{\odot}}$ from asteroseismic modeling. With the orbital inclination ($i = 38.67 \pm 0.29^\circ$) from the astrometric data of Gaia and Hipparcos, we obtain companion masses of $M_2 = 1.76_{-0.042}^{+0.042}~\mathrm{M_{\odot}}$ (radial velocity) and $M_2 = 2.07_{-0.045}^{+0.045}~\mathrm{M_{\odot}}$ (timing variations). The companion's mass exceeds the Chandrasekhar limit and lies in the NS range. Multiple arguments rule out alternatives: the astrometric mass function and the spectral energy distribution, which shows no extra light, together exclude any luminous companion; the mass and lack of Balmer absorption rule out white dwarfs (WDs); the system's age ($1.25$~Gyr) disfavors a double WD; and the mass is too low for a black hole. The wide, low-eccentricity orbit and absence of accretion signatures are consistent with a quiescent NS. SZ~Lyn has the potential to be the first $δ$ Scuti binary with a NS candidate identified through asteroseismology and dynamics, demonstrating the potential of this approach to uncover non-accreting compact objects.

astro-ph.SR↗
arXiv2026

On $\mathcal{F}$-multicolor Turán number of hypergraph graphs

The Ruzsa-Szemerédi $(6,3)$-problem can be equivalently stated as determining the maximum number of edge-disjoint triangles on $n$ vertices such that no triangle is formed by edges from three distinct triangle-copies. Gowers and Janzer extended this problem by establishing an analogous result for complete graphs. A natural generalization of the two results, first introduced by Imolay, Karl, Nagy and Váli, asks for the maximum number of edge-disjoint copies of a graph $F$ on $n$ vertices such that no copy of $G$ is formed by edges originating from distinct $F$-copies. This maximum number, denoted by $ex_F(n,G)$, is called the {\em $F$-multicolor Turán number} of $G$. This paper focuses on the setting of uniform hypergraphs. We first prove that for $k$-uniform hypergraphs $\mathcal{G}$ and $\mathcal{F}$, $ex_{\mathcal{F}}(n,\mathcal{G})=o(n^k)$ if and only if there exists a homomorphism from $\mathcal{G}$ to $\mathcal{F}$. For degenerate case, we show that $ex_{\mathcal{F}}(n,\mathcal{G})=n^{k-o(1)}$ whenever $\mathcal{G}$ contains a $k$-uniform tight triangle. These results extend previous results. We further establish corresponding supersaturation and blowup statements. In the non-degenerate setting, we derive matching lower and upper bounds for $ex_{\mathcal{F}}(n,\mathcal{G})$. We give a necessary and sufficient condition for $ex_{\mathcal{F}}(n,\mathcal{G})$ to fail to attain the upper bound, under the assumption that the extremal graphs for $\mathcal{G}$ are stable. As an application, we refine a result due to Imolay, Karl, Nagy and Váli. Furthermore, we completely characterize $\mathcal{F}$ for which $ex_{\mathcal{F}}(n,\mathcal{G})$ does not attain the upper bound when $\mathcal{G}$ is one of the three special intersecting graphs: Fano plane, extended triangle and $r$-book of $r$-edges with $r=3,4$.

math.CO↗
arXiv2026

Gradient Descent with Projection Finds Over-Parameterized Neural Networks for Learning Low-Degree Polynomials with Nearly Minimax Optimal Rate

We study the problem of learning a low-degree spherical polynomial of degree $k_0 = Θ(1) \ge 1$ defined on the unit sphere in $\RR^d$ by training an over-parameterized two-layer neural network with augmented feature in this paper. Our main result is the significantly improved sample complexity for learning such low-degree polynomials. We show that, for any regression risk $\eps \in (0, Θ(d^{-k_0})]$, an over-parameterized two-layer neural network trained by a novel Gradient Descent with Projection (GDP) requires a sample complexity of $n \asymp Θ( \log(4/δ) \cdot d^{k_0}/\eps)$ with probability $1-δ$ for $δ\in (0,1)$, in contrast with the representative sample complexity $Θ(d^{k_0} \max\set{\eps^{-2},\log d})$. Moreover, such sample complexity is nearly unimprovable since the trained network renders a nearly optimal rate of the nonparametric regression risk of the order $\log({4}/δ) \cdot Θ(d^{k_0}/{n})$ with probability at least $1-δ$. On the other hand, the minimax optimal rate for the regression risk with a kernel of rank $Θ(d^{k_0})$ is $Θ(d^{k_0}/{n})$, so that the rate of the nonparametric regression risk of the network trained by GDP is nearly minimax optimal. In the case that the ground truth degree $k_0$ is unknown, we present a novel and provable adaptive degree selection algorithm which identifies the true degree and achieves the same nearly optimal regression rate. To the best of our knowledge, this is the first time that a nearly optimal risk bound is obtained by training an over-parameterized neural network with a popular activation function (ReLU) and algorithmic guarantee for learning low-degree spherical polynomials. Due to the feature learning capability of GDP, our results are beyond the regular Neural Tangent Kernel (NTK) limit.

stat.ML↗
arXiv2026

Discovering Decoupled Functional Modules in Large Language Models

Understanding the internal functional organization of Large Language Models (LLMs) is crucial for improving their trustworthiness and performance. However, how LLMs organize different functions into modules remains highly unexplored. To bridge this gap, we formulate a functional module discovery problem and propose an Unsupervised LLM Cross-layer MOdule Discovery (ULCMOD) framework that simultaneously disentangles the large set of neurons in the entire LLM into modules while discovering the topics of input samples related to these modules. Our framework introduces a novel objective function and an efficient Iterative Decoupling (IterD) algorithm. Extensive experiments show that our method discovers high-quality, disentangled modules that capture more meaningful semantic information and achieve superior performance in various downstream tasks. Moreover, our qualitative analysis reveals that the discovered modules show semantic coherence, correspond to interpretable specializations, and a clear spatial and hierarchical organization within the LLM. Our work provides a novel tool for interpreting the functional modules of LLMs, filling a critical blank in LLM's interpretability research.

cs.LG↗
arXiv2026

Huawei Cloud Model-as-a-Service on the CloudMatrix384 SuperPod

Scaled-out MoE LLMs and scaled-up SuperPods create new systems challenges for production Model-as-a-Service (MaaS), requiring disaggregation, low-latency communication, and decentralized serving. This report presents xDeepServe, the production serving system behind Huawei Cloud's MaaS offering on CloudMatrix384, a 48-server SuperPod with 384 Ascend 910C chips connected by a high-bandwidth UB fabric and global shared memory. It serves models including DeepSeek, Kimi, GLM, Qwen, and MiniMax, among others. xDeepServe is built around Transformerless, a disaggregated execution architecture that decomposes transformer inference into modular units -- attention, feedforward, and MoE -- and supports disaggregated Prefill-Decode and MoE-Attention deployments. To enable disaggregation, we develop XCCL, a memory-semantic communication layer providing microsecond-level point-to-point and scalable all-to-all primitives, and we extend FlowServe with decentralized DP groups and techniques to mitigate stragglers and synchronization variance. In a peak decoding configuration, xDeepServe reaches 2400 tokens/s per Ascend 910C chip at ~50ms time-per-output-token (TPOT).

cs.DC↗
arXiv2026

Observer-Based Active Fault/Disturbance Compensation Control for Fully Actuated Systems

This paper is concerned with fault/disturbance compensation control for fully actuated systems. In particular, we explore observer-based control, incorporating an active compensation mechanism. First, we propose a novel observer with enhanced design flexibility for the fully actuated system model, enabling simultaneous estimation of system states and exogenous unknown signals, such as faults or disturbances. Then, a nonlinear controller is developed with an active fault or disturbance compensation term, leveraging the fully actuated system approach. The asymptotic stability of both the state estimation error and the closed-loop control system is systematically established. Finally, the feasibility and merits of the proposed method are validated through comparative simulations and experiments.

eess.SY↗
arXiv2026

Sliding Ferroelectricity Induced and Switched Altermagnetism in GaSe-VPSe3-GaSe Sandwiched Heterostructure with Strong Magnetoelectric Effect

Magnetoelectric coupling is vital for exploring fundamental science and driving the development of high-density memory and energy-efficient spintronic devices. Altermagnets, which merge the benefits of ferromagnets and antiferromagnets, pave the way for unprecedented magnetoelectric coupling effects. However, the spin splitting in altermagnets is robustly protected by spin space group symmetry, posing a significant challenge for external manipulation. Here, we propose to utilize the coupling between the layer degree of freedom and the altermagnet to achieve an altermagnetic multiferroic with strong magnetoelectric coupling. In the GaSe-VPSe3-GaSe sandwiched structure, the magnetic order can be switched between altermagnetic and conventional antiferromagnetic by controllably breaking and restoring the combined spatial inversion and time-reversal symmetry using sliding ferroelectricity. Moreover, our systematic investigation of all pathways revealed that the transition from a ferroelectric CB stacking, through an antiferroelectric CC stacking, to a ferroelectric BC stacking is the most favorable, with an energy barrier of only 50.13 meV/f.u.. More importantly, we reveal that the microscopic mechanism of the magnetic phase transition stems from the interlayer covalent bonding of Se-Se or Se-P atomic pairs at the interface. Our findings unveil a new form of magnetoelectric coupling and lay the groundwork for designing miniature information processing and multiferroic memory devices based on altermagnetism.

cond-mat.mtrl-sci↗
arXiv2026

Toward Real-World High-Precision Image Matting and Segmentation

High-precision scene parsing tasks, including image matting and dichotomous segmentation, aim to accurately predict masks with extremely fine details (such as hair). Most existing methods focus on salient, single foreground objects. While interactive methods allow for target adjustment, their class-agnostic design restricts generalization across different categories. Furthermore, the scarcity of high-quality annotation has led to a reliance on inharmonious synthetic data, resulting in poor generalization to real-world scenarios. To this end, we propose a Foreground Consistent Learning model, dubbed as FCLM, to address the aforementioned issues. Specifically, we first introduce a Depth-Aware Distillation strategy where we transfer the depth-related knowledge for better foreground representation. Considering the data dilemma, we term the processing of synthetic data as domain adaptation problem where we propose a domain-invariant learning strategy to focus on foreground learning. To support interactive prediction, we contribute an Object-Oriented Decoder that can receive both visual and language prompts to predict the referring target. Experimental results show that our method quantitatively and qualitatively outperforms SOTA methods.

cs.CV↗
arXiv2026

EMP: Enhance Memory in Data Pruning

Recently, large language and vision models have shown strong performance, but due to high pre-training and fine-tuning costs, research has shifted towards faster training via dataset pruning. Previous methods used sample loss as an evaluation criterion, aiming to select the most "difficult" samples for training. However, when the pruning rate increases, the number of times each sample is trained becomes more evenly distributed, which causes many critical or general samples to not be effectively fitted. We refer to this as Low-Frequency Learning (LFL). In other words, LFL prevents the model from remembering most samples. In our work, we decompose the scoring function of LFL, provide a theoretical explanation for the inefficiency of LFL, and propose adding a memory term to the scoring function to enhance the model's memory capability, along with an approximation of this memory term. Similarly, we explore memory in Self-Supervised Learning (SSL), marking the first discussion on SSL memory. Using contrastive learning, we derive the memory term both theoretically and experimentally. Finally, we propose Enhance Memory Pruning (EMP), which addresses the issue of insufficient memory under high pruning rates by enhancing the model's memory of data, thereby improving its performance. We evaluated the performance of EMP in tasks such as image classification, natural language understanding, and model pre-training. The results show that EMP can improve model performance under extreme pruning rates. For example, in the CIFAR100-ResNet50 pre-training task, with 70\% pruning, EMP outperforms current methods by 2.2\%.

cs.LG↗
arXiv2026

Transverse and Unidirectional Spin Pumping

Conventional spin pumping, driven by magnetization dynamics, is longitudinal since the pumped spin current flows normal to the interface between the ferromagnet and the conductor. We predict \textit{Hall-type/transverse} and \textit{unidirectional} spin pumping into conductors by near-field electromagnetic radiation emitted by, \textit{e.g.}, magnetization dynamics. The joint effect of the electric and magnetic fields results in a pure spin current flowing parallel to the interface, i.e., a Hall-type spin pumping, which is highly efficient due to the strong coupling to the electric field. Such a transverse spin current is unidirectional, with the spatial distribution controlled by the magnetization direction. Our finding reveals a robust approach for generating and manipulating spin currents in future low-dimensional spintronic and orbitronic devices.

cond-mat.mes-hall↗
arXiv2025

Tidally Trapped Two-pole Pulsations Discovered in a Close Binary with a Massive $β$ Cephei Star

Tidally tilted pulsators (TTPs), whose pulsation axis aligns with the binary's semi-major axis, represent a newly established class of oscillators in binary systems. While all previously known TTPs are either $δ$ Scuti or subdwarf B-type stars, their existence has remained unidentified in more massive $β$ Cephei variables. Here, we report the discovery of tidally trapped pulsations in the massive ellipsoidal variable HD~329379, based on photometry from the Transiting Exoplanet Survey Satellite (TESS). Our analysis reveals a $β$ Cephei pulsator in a 2.25-day orbit whose pulsation mode amplitude is strongly modulated with the orbital frequency. Based on our analysis, we concluded that this modulation can be explained by pulsations with significantly larger amplitude near the star's two tidal poles (apsides). We interpret this as a tidally distorted quadrupole pulsation chariacteristiced by trapped two-pole pulsations, with a pulsation axis aligned with the tidal axis. This represents the first identification of such a pulsation mode in a $β$ Cephei star, which differs from single-sided pulsations observed in previous works, marking a rare and important discovery. Our work extends the family of TTPs beyond $δ$ Scuti and subdwarf B-type stars to include more massive $β$ Cephei variables. In particular, the two-pole pulsator HD~329379 stands out as the prototype of a new class of TTPs in massive stars. These results not only provide a new insight to probe the interior structure and evolutionary state for massive stars but also offer a unique opportunity to study the interaction between pulsations and strong tidal distortions.

astro-ph.SR↗
arXiv2025

Asteroseismology and Dynamics Reveal Interior Structure and Coeval Evolution in the Triply Post-Main-Sequence system DG Leo

$δ$ Scuti stars in binary or multiple systems serve as crucial probes for studying stellar pulsation and evolution. However, many such systems are not ideal for asteroseismology due to uncertainties in mass transfer with close companions and the challenges of dynamically measuring all components' physical properties. The triple system DG~Leo, comprising an inner binary and a distant $δ$ Scuti star, is an ideal target due to its well-separated pulsator. By combining new \textit{TESS} photometry with archival spectroscopy, our dynamical analysis shows that the system's three components share similar masses, radii, and luminosities within errors, occupying coincident Hertzsprung--Russell diagram positions, indicative of coeval evolution. By fitting seven observed $δ$ Scuti frequencies through asteroseismic modeling with dynamically constrained theoretical grids, we simultaneously trace the pulsating star's evolution and constrain the triple system's evolutionary stage, with the derived fundamental parameters showing consistency with the dynamical solutions. Our analysis reveals that all three components of DG~Leo are in the post-main-sequence phase, with a system age of $0.7664^{+0.1402}_{-0.1258}$~Gyr. Additionally, the $δ$ Scuti component shows multiple non-radial modes with significant mixed-character frequencies, providing precise constraints on its convective core extent ($R_{\mathrm{cz}}/R = 0.0562^{+0.0137}_{-0.0021}$).

astro-ph.SR↗
arXiv2025

Explicit formulas for the Hattori-Stong theorem and applications

We employ combinatorial techniques to present an explicit formula for the coefficients in front of Chern classes involving in the Hattori-Stong integrability conditions. We also give an evenness condition for the signature of stably almost-complex manifolds in terms of Chern numbers. As an application, it can be showed that the signature of a $2n$-dimensional stably almost-complex manifold whose possibly nonzero Chern numbers being $c_n$ and $c_ic_{n-i}$ is even, which particularly rules out the existence of such structure on rational projective planes. Some other related results and remarks are also discussed in this article.

math.DG↗
arXiv2025

Comparator-Adaptive $Φ$-Regret: Improved Bounds, Simpler Algorithms, and Applications to Games

In the classic expert problem, $Φ$-regret measures the gap between the learner's total loss and that achieved by applying the best action transformation $ϕ\in Φ$. A recent work by Lu et al., [2025] introduces an adaptive algorithm whose regret against a comparator $ϕ$ depends on a certain sparsity-based complexity measure of $ϕ$, (almost) recovering and interpolating optimal bounds for standard regret notions such as external, internal, and swap regret. In this work, we propose a general idea to achieve an even better comparator-adaptive $Φ$-regret bound via much simpler algorithms compared to Lu et al., [2025]. Specifically, we discover a prior distribution over all possible binary transformations and show that it suffices to achieve prior-dependent regret against these transformations. Then, we propose two concrete and efficient algorithms to achieve so, where the first one learns over multiple copies of a prior-aware variant of the Kernelized MWU algorithm of Farina et al., [2022], and the second one learns over multiple copies of a prior-aware variant of the BM-reduction [Blum and Mansour, 2007]. To further showcase the power of our methods and the advantages over Lu et al., [2025] besides the simplicity and better regret bounds, we also show that our second approach can be extended to the game setting to achieve accelerated and adaptive convergence rate to $Φ$-equilibria for a class of general-sum games. When specified to the special case of correlated equilibria, our bound improves over the existing ones from Anagnostides et al., [2022a,b]

cs.LG↗