Yunjie Ye | Searcharxiv

arXiv2026

A Thermodynamic-Limit Pinning Criterion for Two-Dimensional Structural Superlubricity

Incommensurability and elastic reconstruction do not by themselves define a structurally superlubric phase. We define fully sliding and pinned zero-temperature phases by $\limsup_{A\to\infty}τ_{\rm dep}^{\max}(A)=0$ and $\liminf_{A\to\infty}τ_{\rm dep}^{\min}(A)>0$, respectively; $Λ_n=|V_n|G_{n,i}[D_{\rm rel}^{-1}(\mathbf q_n)]_{ij}G_{n,j}$ measures only reconstruction susceptibility. Translational covariance then proves that a clean, smooth, infinite moiré continuum can reconstruct without acquiring a bulk sliding barrier. We restore atomic sampling in a two-dimensional discrete model of graphene/hBN and test both a diffusion quantum Monte Carlo first-star potential and a 15-harmonic Leven potential across three rational approximants and five directions. No physical-coupling equilibrium or metastable barrier is resolved. The Leven spectrum raises the largest tested $Λ$ from $0.142$ to $0.212$, while artificial scaling through $Λ=1$ reaches uncontrolled strain before a size-independent threshold appears. The tested zero-temperature in-plane models are therefore consistent with an elastically relaxed sliding regime; $Λ=1$ is a reconstruction scale, not a static phase criterion.

cond-mat.mtrl-sci↗

arXiv2023

Exploiting Counter-Examples for Active Learning with Partial labels

This paper studies a new problem, \emph{active learning with partial labels} (ALPL). In this setting, an oracle annotates the query samples with partial labels, relaxing the oracle from the demanding accurate labeling process. To address ALPL, we first build an intuitive baseline that can be seamlessly incorporated into existing AL frameworks. Though effective, this baseline is still susceptible to the \emph{overfitting}, and falls short of the representative partial-label-based samples during the query process. Drawing inspiration from human inference in cognitive science, where accurate inferences can be explicitly derived from \emph{counter-examples} (CEs), our objective is to leverage this human-like learning pattern to tackle the \emph{overfitting} while enhancing the process of selecting representative samples in ALPL. Specifically, we construct CEs by reversing the partial labels for each instance, and then we propose a simple but effective WorseNet to directly learn from this complementary pattern. By leveraging the distribution gap between WorseNet and the predictor, this adversarial evaluation manner could enhance both the performance of the predictor itself and the sample selection process, allowing the predictor to capture more accurate patterns in the data. Experimental results on five real-world datasets and four benchmark datasets show that our proposed method achieves comprehensive improvements over ten representative AL frameworks, highlighting the superiority of WorseNet. The source code will be available at \url{https://github.com/Ferenas/APLL}.

cs.LG↗