arXiv subjects

Xinye Li

Publications and source records attributed to Xinye Li.

13 recordsLinked to original sources

arXiv2026

Existence and strong-field asymptotics of skyrmions in a fourth-order model of frustrated ferromagnets

We study a fourth-order variational model for two-dimensional frustrated ferromagnets with competing exchange interactions and an applied magnetic field of strength $H>0$. For every $H>1/4$, we prove that the energy admits minimizers in the topological classes $Q=\pm1$ and that every minimizing sequence is precompact in $H^2$ modulo translations. The main difficulty is that spectral coercivity degenerates as $H\downarrow1/4$. Using a Helmholtz circle-mean identity, we prove that the residual energy of every nonzero-degree configuration has a uniform positive lower bound, even at the degenerate endpoint. Together with a sphere-valued $H^2$-splitting construction, this provides a threshold-stable binding inequality and yields compactness throughout the coercive regime. We also identify $H=1/4$ as the sharp spectral threshold. Below it the energy is unbounded from below, whereas at the threshold nonzero-degree configurations retain a positive energy barrier and degree-zero Weyl sequences lose compactness. Finally, in the strong-field regime, rescaled minimizers approach those minimizers of the limiting functional that maximize the Dirichlet energy, while topological-charge and normalized-energy measures concentrate on the $H^{-1/4}$-scale.

math.AP↗

arXiv2026

ForgeWM: Progressive Causal Training for Few-Step Action-Conditioned Video World Models

Action-conditioned video world models require low-latency causal generation and reliable responses to game-native controls. Although causal distillation enables one- or few-step video synthesis, extending it to interactive world models remains challenging, as discrete keyboard states and continuous mouse motion must remain aligned with temporally compressed latent chunks during causal training and autoregressive rollout. We introduce ForgeWM, a progressive framework that transforms a bidirectional action-conditioned video generator into efficient few-step world models through domain adaptation, teacher-forced causal training, causal consistency distillation, and on-policy distribution matching with a bidirectional teacher. The resulting budget-specialized students operate at steady-state denoising budgets of 1, 2, and 4 steps. ForgeWM further supports a dual-path deployment protocol combining latency-critical interaction with optional replay-time refinement, where the one-step student re-noises and refines its saved draft. On paired Minecraft trajectories, ForgeWM leads the evaluated systems in Imaging Quality, reference-aligned motion-profile agreement, action-sign accuracy, and mouse-control accuracy, while achieving the lowest reference LPIPS; the same four-stage recipe transfers to gamepad-controlled FPS gameplay. Replay-time refinement matches four-step reference quality while remaining roughly three times closer to the experienced trajectory than regeneration from noise. These results demonstrate ForgeWM's effectiveness for controllable few-step video generation.

cs.CV↗

arXiv2026

Measure, Don't Optimize: Forecasting Recovery in LLM Unlearning

Prior white-box studies show that large language models can retain latent traces of target knowledge after unlearning, even when the knowledge is no longer expressed in their outputs. However, existing audits remain limited to one-off diagnostics: it is unclear whether these residual signals can predict future recovery under continued training or serve as reliable optimization targets. Resolving this gap is essential to determine whether internal auditing can move beyond post-hoc evaluation toward proactive risk monitoring and safer unlearning. We propose J-Access, an inference-time audit that uses the Jacobian lens to map intermediate representations into vocabulary space and measures how often target concepts remain accessible along the model's output pathway. We hypothesize that residual accessibility reflects recovery susceptibility: knowledge that remains closer to the output pathway requires less fine-tuning to restore, leading to faster recovery. We audit 398 public unlearned models spanning eight unlearning methods. We find that: (1) most unlearned models retain access above the retain-only gold level; (2) pre-attack accessibility predicts recovery speed and extent at the model level, but cannot identify which specific facts will be recovered; and (3) directly minimizing J-Access does not promote genuine deletion. Instead, the model learns to hide knowledge from the audit, producing lower audit scores but greater post-attack recovery. These findings position J-Access as a model-level diagnostic for assessing residual susceptibility in unlearned models. We argue internal audits should serve as an independent diagnostic dimension in unlearning evaluation, and should not be converted into optimization targets without validation.

cs.CL↗

arXiv2025

Cognitive Alpha Mining via LLM-Driven Code-Based Evolution

Discovering effective predictive signals, or "alphas," from financial data with high dimensionality and extremely low signal-to-noise ratio remains a difficult open problem. Despite progress in deep learning, genetic programming, and, more recently, large language model (LLM)-based factor generation, existing approaches still explore only a narrow region of the vast alpha search space. Neural models tend to produce opaque and fragile patterns, while symbolic or formula-based methods often yield redundant or economically ungrounded expressions that generalize poorly. Although different in form, these paradigms share a key limitation: none can conduct broad, structured, and human-like exploration that balances logical consistency with creative leaps. To address this gap, we introduce the Cognitive Alpha Mining Framework (CogAlpha), which combines code-level alpha representation with LLM-driven reasoning and evolutionary search. Treating LLMs as adaptive cognitive agents, our framework iteratively refines, mutates, and recombines alpha candidates through multi-stage prompts and financial feedback. This synergistic design enables deeper thinking, richer structural diversity, and economically interpretable alpha discovery, while greatly expanding the effective search space. Experiments on 5 stock datasets from 3 stock markets demonstrate that CogAlpha consistently discovers alphas with superior predictive accuracy, robustness, and generalization over existing methods. Our results highlight the promise of aligning evolutionary optimization with LLM-based reasoning for automated and explainable alpha discovery.

cs.CL↗

arXiv2025

Record High Polarization at 2V and Imprint-free operation in Superlattice HfO2-ZrO2 by Proper Tuning of Ferro and Antiferroelectricity

Neuromorphic computing, inspired by biological intelligence, offers a pathway to revolutionize artificial intelligence (AI) by unifying memory and processing in an energy-efficient, sustainable framework for data-intensive tasks. Ferroelectric (FE) materials have emerged as promising candidates for implementing artificial synapses, yet achieving low-voltage operation in CMOS back-end compatible devices remains a major challenge. In this work, we demonstrate that proper tuning of ferro and antiferroelectric phases in HfO2-ZrO2(HZO) superlattice based capacitors can lead to imprint-free switching with record switchable polarization (2Pr) of 76 micro Coulomb cm-2 under an external field of only 2 MV cm-1. The sizable remanent polarization of the superlattice HZO further enables linear potentiation and depression with an on to off ratio of 20 within a 3 MV cm-1 bias window. Under pulsed operation, the devices show robust endurance, either maintaining polarization with less than 10 per cent degradation up to 10^8 cycles or surviving beyond 10^9 cycles with recoverable fatigue. By elucidating two distinct fatigue mechanisms, this work highlights strategies for optimizing FE devices to meet the stringent demands of neuromorphic training applications.

cond-mat.mtrl-sci↗

arXiv2025

ScEdit: Script-based Assessment of Knowledge Editing

Knowledge Editing (KE) has gained increasing attention, yet current KE tasks remain relatively simple. Under current evaluation frameworks, many editing methods achieve exceptionally high scores, sometimes nearing perfection. However, few studies integrate KE into real-world application scenarios (e.g., recent interest in LLM-as-agent). To support our analysis, we introduce a novel script-based benchmark -- ScEdit (Script-based Knowledge Editing Benchmark) -- which encompasses both counterfactual and temporal edits. We integrate token-level and text-level evaluation methods, comprehensively analyzing existing KE techniques. The benchmark extends traditional fact-based ("What"-type question) evaluation to action-based ("How"-type question) evaluation. We observe that all KE methods exhibit a drop in performance on established metrics and face challenges on text-level metrics, indicating a challenging task. Our benchmark is available at https://github.com/asdfo123/ScEdit.

cs.CL↗

arXiv2025

LLMSR@XLLM25: An Empirical Study of LLM for Structural Reasoning

We present Team asdfo123's submission to the LLMSR@XLLM25 shared task, which evaluates large language models on producing fine-grained, controllable, and interpretable reasoning processes. Systems must extract all problem conditions, decompose a chain of thought into statement-evidence pairs, and verify the logical validity of each pair. Leveraging only the off-the-shelf Meta-Llama-3-8B-Instruct, we craft a concise few-shot, multi-turn prompt that first enumerates all conditions and then guides the model to label, cite, and adjudicate every reasoning step. A lightweight post-processor based on regular expressions normalises spans and enforces the official JSON schema. Without fine-tuning, external retrieval, or ensembling, our method ranks 5th overall, achieving macro F1 scores on par with substantially more complex and resource-consuming pipelines. We conclude by analysing the strengths and limitations of our approach and outlining directions for future research in structural reasoning with LLMs. Our code is available at https://github.com/asdfo123/LLMSR-asdfo123.

cs.CL↗

arXiv2024

Towards Improved Polarization Uniformity in Ferroelectric Hf$_{0.5}$Zr$_{0.5}$O$_2$ Devices within Back End of Line Thermal Budget for Memory and Neuromorphic Applications

Thin film ferroelectric devices with ultralow power operation, non-volatile data retention and fast and reliable switching are attractive for non-volatile memory and as synaptic weight elements. However, low thermal budget ferroelectric oxides suffer from crystalline inhomogeneity and defects that makes their large-scale circuit integration challenging. Here, we report on the thermally engineered way to induce wafer-scale homogeneity in Hf$_{0.5}$Zr$_{0.5}$O$_2$ capacitors that can lead to high device reliability making their integration possible in ultralow power memory and neuromorphic computing hardware.

physics.app-ph↗

arXiv2024

Designing high endurance Hf0.5Zr0.5O2 capacitors through engineered recovery from fatigue for non-volatile ferroelectric memory and neuromorphic hardware

Heavy computational demands from artificial intelligence (AI) leads the research community to explore the design space for functional materials that can be used for high performance memory and neuromorphic computing hardware. Novel device technologies with specially engineered properties are under intense investigation to revolutionize information processing with brain-inspired computing primitives for ultra energy-efficient implementation of AI and machine learning tasks. Ferroelectric memories with ultra-low power and fast operation, non-volatile data retention and reliable switching to multiple polarization states promises one such option for non-volatile memory and synaptic weight elements in neuromorphic hardware. For quick adaptation of industry, new materials need complementary metal oxide semiconductor (CMOS) process compatibility which brings a whole new set of challenges and opportunities for advanced materials design. In this work, we report on designing of back-end-of-line compatible ferroelectric Hf0.5Zr0.5O2 capacitors that are capable of recovery from fatigue multiple times reaching 2Pr > 40 microC cm-2 upon each retrieval. Our results indicate that with specifically engineered material stack and annealing protocols, it is possible to reach endurance exceeding 10^9 cycles at room temperature that can lead to ultralow power ferroelectric non-volatile memory components or synaptic weight elements compatible with online training or inference tasks for neuromorphic computing.

cond-mat.mtrl-sci↗

arXiv2023

Well-posedness and stability for a class of fourth-order nonlinear parabolic equations

In this paper we examine well-posedness for a class of fourth-order nonlinear parabolic equation $\partial_t u + (-\Delta)^2 u = \nabla \cdot F(\nabla u)$, where $F$ satisfies a cubic growth conditions. We establish existence and uniqueness of the solution for small initial data in local BMO spaces. In the cubic case $F(\xi) = \pm \lvert \xi \rvert^2 \xi$ we also examine the large time behaivour and stability of global solutions for arbitrary and small initial data in VMO, respectively.

math.AP↗

arXiv2020

Lattice solutions in a Ginzburg-Landau model for a chiral magnet

We examine micromagnetic pattern formation in chiral magnets, driven by the competition of Heisenberg exchange, Dzyaloshinskii-Moriya interaction, easy-plane anisotropy and thermodynamic Landau potentials. Based on equivariant bifurcation theory we prove existence of lattice solutions branching off the zero magnetization state and investigate their stability. We observe in particular the stabilization of quadratic vortex-antivortex lattice configurations and instability of hexagonal skyrmion lattice configurations, and we illustrate our findings by numerical studies.

math-ph↗

arXiv2017

Stability of axisymmetric chiral skyrmions

We examine topological solitons in a minimal variational model for a chiral magnet, so-called chiral skyrmions. In the regime of large background fields, we prove linear stability of axisymmetric chiral skyrmions under arbitrary perturbations in the energy space, a long-standing open question in physics literature. Moreover, we show strict local minimality of axisymmetric chiral skyrmions and nearby existence of moving soliton solution for the Landau-Lifshitz-Gilbert equation driven by a small spin transfer torque.

math.CA↗

arXiv2016

The stability of fixed points for a Kuramoto model with Hebbian interactions

We consider a variation of the Kuramoto model with dynamic coupling, where the coupling strengths are allowed to evolve in response to the phase difference between the oscillators, a model first considered by Ha, Noh and Park. In particular we study the stability of fixed points for this model. We demonstrate a somewhat surprising fact: namely that the fixed points of this model, as well as their stability, can be completely expressed in terms of the fixed points and stability of the analogous classical Kuramoto problem where the coupling strengths are fixed to a constant (the same for all edges). In particular for the "all-to-all" network, where the underlying graph is the complete graph, the problem reduces to the problem of understanding the fixed points and stability of the all-to-all Kuramoto model with equal edge weights, a problem that has been completely solved.

math.DS↗