arXiv subjects

Hongze Li

Publications and source records attributed to Hongze Li.

At least 19 recordsLinked to original sources

arXiv2026

Balanced electron and phonon heat transport in metallic $\varepsilon$-TaN

Most materials with high thermal conductivity belong to one of two classes: metals, where heat is carried predominantly by electrons, and insulators, where heat transport is dominated by the phonon contribution. Materials that combine substantial electronic thermal conductivity and lattice thermal conductivity are rare, because the mechanisms that favor electron transport typically suppress phonon transport, and vice versa. Here, we report the theoretical prediction and experimental realization of such a material, metallic $\varepsilon$-TaN. Our calculations predict a total thermal conductivity at room-temperature of 273$\pm$5Wm$^{-1}$K$^{-1}$ in single crystals and 145$\pm$5Wm$^{-1}$K$^{-1}$ in polycrystals with 0.5$\mu$m grains, with an unusually large lattice contribution (79%) for a metal. The latter value is in agreement with our local transient thermoreflectance measurements on polycrystalline samples yielding $\sim$130Wm$^{-1}$K$^{-1}$. We show that the balanced electronic and lattice thermal conductivities of $\varepsilon$-TaN originate from a combination of large Fermi velocity and small Fermi density of states on the electron side, and large speed of sound and wide phonon gap on the lattice side.

cond-mat.mtrl-sci↗

arXiv2026

ViCrop-Det: Spatial Attention Entropy Guided Cropping for Training-Free Small-Object Detection

Transformer-based architectures have established a dominant paradigm in global semantic perception; however, they remain fundamentally constrained by the profound spatial heterogeneity inherent in natural images. Specifically, the imposition of a uniform global receptive field across regions of varying information density inevitably leads to local feature degradation, particularly in dense conflict zones populated by microscopic targets. To address this mechanistic limitation, we propose ViCrop-Det, a training-free inference framework that introduces adaptive spatial trust region shrinkage. Inspired by the use of attention entropy in anomaly segmentation, ViCrop-Det leverages the detection decoder's cross-attention distribution as an endogenous probe. By utilizing Spatial Attention Entropy (SAE) to heuristically evaluate local spatial ambiguity, the framework executes dynamic spatial routing, allocating a fixed computational budget exclusively to regions exhibiting both high target saliency and high cognitive uncertainty. By shrinking the spatial trust region and injecting high-frequency localized observations, ViCrop-Det actively resolves spatial ambiguity and recovers fine-grained features without requiring architectural modifications. Extensive evaluations on VisDrone and DOTA-v1.5 demonstrate that ViCrop-Det yields competitive performance enhancements, consistently adding +1-3 mAP@50 to RT-DETR-R50 and Deformable DETR with a marginal 20-23\% latency overhead. On MS COCO, $AP_{S}$ improves while $AP_{M}/AP_{L}$ remains stable, indicating precise fine-scale refinement without compromising the global spatial prior. Under compute-matched settings, our adaptive routing strategy comprehensively surpasses uniform slicing baselines, achieving a highly optimized accuracy-speed trade-off.

cs.CV↗

arXiv2026

Dynamics of spin glasses in two dimensions

Spin glass dynamics is a strong function of spatial dimensionality $D$. The lower critical dimension is close to 2.5, so that, in two dimensions, the condensation temperature $T_\text{g}=0$, and only fluctuations are present at finite temperatures. However, by using thin film multilayers, one can explore the dynamics in both $D=3$ and $D=2$ dimensions. Spin glass thin film multilayers transition from $D=3$ dynamics at short to intermediate times to $D = 2$ dynamics at long times. Correlation lengths of CuMn 4.5 nm multilayers at long times are shown to be grow more rapidly in $D=2$ as compared to $D=3$, and for the longest measurement time, experimentally reach equilibrium in qualitative agreement with simulations.

cond-mat.dis-nn↗

arXiv2025

Least zero of pairs of additive cubic equations

An effective upper bound is established for the least non-trivial integer solution to the system of cubic forms \[ \begin{cases} F = c_{1}x_1^3 + c_{2}x_2^3 + \cdots + c_{n}x_n^3 = 0, \\ G = d_{1}x_1^3 + d_{2}x_2^3 + \cdots + d_{n}x_n^3 = 0, \end{cases} \] under the "$M$-good" condition for $n \ge 16$, where $c_{1}, \dots, c_{n}$ and $d_{1}, \dots, d_{n}$ are integers. Additionally, a range is derived for the probability that randomly selected simultaneous equations satisfy the $M$-good condition.

math.NT↗

arXiv2025

FedSDAF: Leveraging Source Domain Awareness for Enhanced Federated Domain Generalization

Traditional Federated Domain Generalization (FedDG) methods focus on learning domain-invariant features or adapting to unseen target domains, often overlooking the unique knowledge embedded within the source domain, especially in strictly isolated federated learning environments. Through experimentation, we discovered a counterintuitive phenomenon: features learned from a complete source domain have superior generalization capabilities compared to those learned directly from the target domain. This insight leads us to propose the Federated Source Domain Awareness Framework (FedSDAF), the first systematic approach to enhance FedDG by leveraging source domain-aware features. FedSDAF employs a dual-adapter architecture that decouples "local expertise" from "global generalization consensus." A Domain-Aware Adapter, retained locally, extracts and protects the unique discriminative knowledge of each source domain, while a Domain-Invariant Adapter, shared across clients, builds a robust global consensus. To enable knowledge exchange, we introduce a Bidirectional Knowledge Distillation mechanism that facilitates efficient dialogue between the adapters. Extensive experiments on four benchmark datasets (OfficeHome, PACS, VLCS, and DomainNet) show that FedSDAF significantly outperforms existing FedDG methods. The source code is available at https://github.com/pizzareapers/FedSDAF.

cs.LG↗

arXiv2024

Nature of temperature chaos in spin glasses

Temperature chaos (TC) in spin glasses has been claimed to exist no matter how small the temperature change, $\Delta T$. However, experimental studies have exhibited a finite value of $\Delta T$ for a transition to TC. This paper explores the onset of TC with much higher resolution than before and over a larger temperature range. We find that TC is always present, though small at the smallest $\Delta T$ that we can reliably measure. However, it grows rapidly as $\Delta T$ increases, the region of rapid growth coinciding with the $\Delta T$ predicted from renormalization group arguments and observed experimentally. We are able to transcend the full range of TC, from the completely reversible state to one that is maximally decorrelated from the initially prepared state.

cond-mat.dis-nn↗

arXiv2024

Least zero of a cubic form

An explicit upper bound is established for the least non-trivial integer zero of an arbitrary cubic form $C \in \mathbb{Z}[X_1,...,X_n],$ provided that $n \geq 14.$

math.NT↗

arXiv2024

Exotic magnetism in perovskite KOsO3

A new perovskite KOsO3 has been stabilized under high-pressure and high temperature conditions. It is cubic at 500 K (Pm-3m) and undergoes subsequent phase transitions to tetragonal at 320 K (P4/mmm) and rhombohedral (R-3m) at 230 K as shown from refining synchrotron X-ray powder diffraction (SXRD) data. The larger orbital overlap integral and the extended wavefunction of 5d electrons in the perovskite KOsO3 allow to explore physics from the regime where Mott and Hund's rule couplings dominate to the state where the multiple interactions are on equal footing. We demonstrate an exotic magnetic ordering phase found by neutron powder diffraction along with physical properties via a suite of measurements including magnetic and transport properties, differential scanning calorimetry, and specific heat, which provide comprehensive information for a system at the crossover from localized to itinerant electronic behavior.

cond-mat.str-el↗

arXiv2024

Synthesis of epitaxial magnetic pyrochlore heterojunctions

The synthesis of stoichiometric and epitaxial pyrochlore iridate thin films presents significant challenges yet is critical for unlocking experimental access to novel topological and magnetic states. Towards this goal, we unveil an in-situ two-stage growth mechanism that facilitates the synthesis of high-quality oriented pyrochlore iridate thin films. The growth starts with the deposition of a pyrochlore titanate as an active iso-structural template, followed by the application of an in-situ solid phase epitaxy technique in the second stage to accomplish the formation of single crystalline, large-area films. This novel protocol ensures the preservation of stoichiometry and structural homogeneity, leading to a marked improvement in surface and interface qualities over previously reported methods. The success of this synthesis approach is attributed to the application of directional laser-heat annealing, which effectively reorganizes the continuous random network of ions into a crystalline structure, as evidenced by our comprehensive analysis of the growth kinetics. This new synthesis approach advances our understanding of pyrochlore iridate film fabrication and opens a new perspective for investigating their unique physical properties.

cond-mat.str-el↗

arXiv2023

ALUM: Adversarial Data Uncertainty Modeling from Latent Model Uncertainty Compensation

It is critical that the models pay attention not only to accuracy but also to the certainty of prediction. Uncertain predictions of deep models caused by noisy data raise significant concerns in trustworthy AI areas. To explore and handle uncertainty due to intrinsic data noise, we propose a novel method called ALUM to simultaneously handle the model uncertainty and data uncertainty in a unified scheme. Rather than solely modeling data uncertainty in the ultimate layer of a deep model based on randomly selected training data, we propose to explore mined adversarial triplets to facilitate data uncertainty modeling and non-parametric uncertainty estimations to compensate for the insufficiently trained latent model layers. Thus, the critical data uncertainty and model uncertainty caused by noisy data can be readily quantified for improving model robustness. Our proposed ALUM is model-agnostic which can be easily implemented into any existing deep model with little extra computation overhead. Extensive experiments on various noisy learning tasks validate the superior robustness and generalization ability of our method. The code is released at https://github.com/wwzjer/ALUM.

cs.LG↗

arXiv2022

A new type of degenerate poly-Euler polynomials

Many mathematicians have been studying various degenerate versions of special polynomials and numbers in some arithmetic and combinatorial aspects. Our main focus here is a new type of degenerate poly-Euler polynomials and numbers. This focus stems from their nascent importance for applications in combinatorics, number theory and in other aspects of applied mathematics. we construct a new type of degenerate poly-Euler polynomials by using the degenerate polylogarithm functions. We also show several combinatorial identities related to this polynomials and numbers.

math.CO↗

arXiv2019

The Green-Tao theorem for Piatetski-Shapiro primes

Let $m\geq 3$. Suppose that $$ 1-2^{-2^{m^24^m}}<γ<1. $$ Then the set $$ \{p\text{ prime}:\, p=[n^{\frac1γ}]\text{ for some }n\in{\mathbb N}\} $$ contains infinitely many non-trivial $m$-term arithmetic progressions.

math.NT↗

arXiv2018

A Visual Query System for Scholar Networks

Large scholar networks is quite popular in the academic domain, like Aminer. It offers to display the academic social network, including profile search, expert finding, conference analysis, course search, sub-graph search, topic browser, academic ranks and user management. Usually the search results are listed as items, while the relations among them are hidden to the users. Visualization is a feasible way to help users explore the hidden relations and discover more useful information. This article aim to visualize the search results in Aminer in a more user-friendly way and help them better utilize the tool. We provided three different designs to visualize the results and tested them in user study. The empirical results of our research show that the designed graphs help users better understand the area they intend to know and make their search more effective.

cs.IR↗

arXiv2016

The large $k$-term progression-free sets in $\mathbb{Z}_q^n$

Let $k$ and $n$ be fixed positive integers. For each prime power $q\geqslant k\geqslant 3$, we show that any subset $A\subseteq \mathbb{Z}_q^n$ free of $k$-term arithmetic progressions has size $|A|\leqslant c_k(q)^n$ with a constant $c_k(q)$ that can be expressed explicitly in terms of $k$ and $q$. As a consequence, we can take $c_k(q)=0.8415q$ for sufficiently large $q$ and arbitrarily fixed $k\geq 3$.

math.NT↗

arXiv2016

Small gaps between the Piatetski-Shapiro primes

Suppose that $1<c<9/8$. For any $m\geq 1$, there exist infinitely many $n$ such that $$ \{[n^c],\ [(n+1)^c],\ \ldots,\ [(n+k_0)^c]\} $$ contains at least $m+1$ primes, if $k_0$ is sufficiently large (only depending on $m$).

math.NT↗

arXiv2015

Bounded gaps between primes of the special form

For each $m\geq 1$, there exist infinitely many primes $p_1<p_2<\ldots<p_{m+1}$ such that $p_{m+1}-p_1=O(m^4e^{8m})$ and $p_j+2$ has at most $\frac{16m}{\log 2}+\frac{5\log m}{\log 2}+37$ prime divisors for each $j$.

math.NT↗

arXiv2013

On sums of sparse prime subsets

For arbitrary $c_0>0$, if $A$ is a subset of the primes less than $x$ with cardinality $δx (\log x)^{-1}$ with $δ\geq (\log x)^{-c_0}$, then there exists a positive constant $c$ such that the cardinality of $A+A$ is larger than $c\, δx (\log\log x)^{-1}$.

math.NT↗

arXiv2012

On sums of subsets of Chen primes

In this paper we show that if $A$ is a subset of Chen primes with positive relative density $α$, then $A+A$ must have positive upper density at least $cαe^{-c^\prime\log(1/α)^{2/3}(\log\log(1/α))^{1/3}}$ in the natural numbers.

math.NT↗