arXiv subjects

Jiayu Liang

Publications and source records attributed to Jiayu Liang.

10 recordsLinked to original sources

arXiv2026

Unveiling Nanoscale Surface Damage Dynamics in Swift Heavy Ion Irradiated Gallium Nitride

This work systematically unveils the nanoscale surface damage dynamics in gallium nitride by investigating the atomistic mechanisms of hillock formation. The results identify two distinct hillock morphologies dependent on electronic energy loss (Se) values. Bell-shaped hillocks form under 18.2 keV/nm Kr irradiation, whereas crater-rim hillocks with central holes emerge under 40.2 keV/nm Ta irradiation. Microstructural analysis reveals that Ga-rich hillocks are accompanied by the generation of metastable zincblende nanodomains. These nanodomains preferentially aggregate around the periphery or sidewalls of the hillocks and exhibit a high spatial correlation with screw dislocations. Further temperature-dependent studies indicate that elevated temperatures significantly enlarge the overall dimensions of the hillock structures without altering their fundamental morphologies. Notably, under Ta irradiation above 1200 K, the high temperatures drastically reduce the viscosity and surface tension of liquid gallium. This enhanced fluidity of the transient molten phase promotes the formation of penetrating nanochannels.

cond-mat.mtrl-sci↗

arXiv2026

Anisotropic Core-Shell Swift Heavy Ion Tracks in beta-Ga2O3

Swift heavy ion (SHI) irradiation generates nanoscale ion tracks through intense electronic excitation, yet the microscopic mechanisms governing their morphology and phase stability in low symmetry oxides remain poorly understood. Here, a multiscale atomistic simulation framework is used to investigate the formation and recovery of SHI-induced tracks in monoclinic $β$-Ga2O3 over a wide range of electronic energy losses (Se) and crystallographic orientations. A sequence of distinct structural responses is identified with increasing Se: (i) complete lattice recovery at low Se; (ii) recrystallization into a metastable $γ$-Ga2O3 phase at intermediate Se; and (iii) the formation of core-shell ion tracks at high Se, consisting of an amorphous core surrounded by a recrystallized $γ$-phase shell. Despite the essentially isotropic initial energy deposition, the final ion-track morphology exhibits pronounced crystallographic anisotropy, governed by orientation-dependent recovery dynamics. The superior recrystallization along the [010] direction is attributed to its exceptionally high elastic stiffness. Notably, SHI irradiation perpendicular to the (100) plane induces a more severe structural response at low Se ($\le$ 10 keV/nm), however, at higher Se, it yields a smaller residual ion track compared to the other orientations. The simulated ion-track sizes show excellent quantitative agreement with the available experimental measurements over a wide range of Se values. These findings establish a unified atomic-scale picture of core-shell track formation and anisotropic recovery in $β$-Ga2O3.

cond-mat.mtrl-sci↗

arXiv2026

Thought-Aware KV Cache Compaction for Reasoning via Adaptive Attention Matching

Reasoning language models generate lengthy chain-of-thought (CoT) sequences whose key-value (KV) cache grows linearly and becomes a memory bottleneck during decoding. Existing compaction methods treat reasoning trajectories as flat token sequences and apply uniform compression, ignoring the hierarchical structure of CoT reasoning where different steps vary drastically in importance. We propose \textbf{Thought-Aware Attention Matching (TAM)}, which exploits this structure through three mechanisms: (i)~thought segmentation that decomposes the trajectory into reasoning blocks, (ii)~adaptive budget allocation that assigns compression budget based on each segment's importance and size, and (iii)~pivotal token protection that preserves high-attention reasoning anchors. We prove that the allocation rule is optimal under a convex error model and that cumulative error under sequential compaction remains bounded. Experiments on AIME 2024 and MATH-500 with Qwen3-4B show that TAM improves accuracy over uniform compaction at the same memory footprint, with periodic compaction bounding peak memory to 3.1--3.2\,GB (a 65\% reduction) while maintaining competitive accuracy.

cs.CL↗

arXiv2026

ExTax: Explainable Disinformation Detection via Persuasion, Emotion, and Narrative Role Taxonomies

The democratization of LLMs has accelerated the generation and circulation of highly fluent disinformation, making traditional syntax-semantic verification increasingly insufficient. Such deception rarely relies solely on surface-level falsity; instead, it often combines persuasive rhetoric, emotional manipulation, and narrative role construction to influence readers' interpretations through multiple cognitive pathways. However, existing detectors typically emphasize isolated signals -- such as syntax, external knowledge, persuasion, or affective cues -- and therefore struggle to capture the multi-faceted manipulative intents underlying disinformation or provide human-auditable explanations. To address this gap, we present \textbf{ExTax}, a taxonomy-aligned framework for explainable disinformation detection. ExTax unifies persuasive rhetoric, emotional manipulation, and narrative roles into a 17-dimensional taxonomic space, covering 6 persuasive-rhetoric strategies, 5 emotional-manipulation methods, and 6 narrative-role categories. It elicits attributes from multiple frontier LLMs, reconciles their disagreements through Entropy-driven Dynamic Label Smoothing, and fuses the resulting taxonomic representations with contextual encodings via Heterogeneous Multi-Head Attention, grounding each prediction in an interpretable manipulation profile. Across five cross-domain and cross-genre benchmarks, ExTax achieves an overall Macro $F_1$ of $0.8456$, outperforming state-of-the-art deep learning and LLM-based baselines. It also remains robust under severe genre imbalance, where the strongest deep baseline degrades from $0.9454$ to $0.6194$.

cs.CL↗

arXiv2026

FinHarness: An Inline Lifecycle Safety Harness for Finance LLM Agents

Finance LLM agents must simultaneously block prompt-induced unauthorized actions and approve legitimate multi-step business workflows. However, boundary filters often miss irreversible mid-trajectory tool calls, while post-hoc LLM judges perform auditing only after termination -- too late for intervention and at a computational cost that scales linearly with trace length. We present FinHarness, an inline safety harness that wraps a finance agent end-to-end with three components: a Query Monitor that fuses single-turn intent with cross-turn drift, a Tool Monitor that evaluates each prospective tool call, and a Cascade module that integrates per-step risk and adaptively routes verification between a lightweight and an advanced-tier LLM judge. Fired risk factors are re-injected into the agent input as ex-ante evidence, enabling the agent to refuse, re-plan, or approve on its own. On FinVault, routed FinHarness cuts ASR from 38.3% to 15.0% while largely preserving benign approval ($41.1\% \to 39.3\%$), and uses $4.7\times$ fewer advanced-judge calls than an always-advanced ablation.

cs.CL↗

arXiv2026

Atomistic Mechanisms of Temperature-Dependent Ion Track Formation in Gallium Nitride under Swift Heavy Ion Irradiation

The radiation tolerance of gallium nitride under extreme conditions is critical for its deployment in next-generation electronic and optoelectronic devices, yet the microscopic mechanisms governing swift heavy ion induced damage at elevated temperatures remain poorly understood. Therefore, this study employs a coupled approach including the two-temperature model and molecular dynamics simulations to resolve the entire processes of ion track generation induced by swift heavy ions irradiation across a wide temperature range. A temperature-driven morphological transition of ion tracks, evolving from discontinuous segments to continuous tracks composed of isolated nanobubbles, and ultimately to fully continuous channels is observed. Under lower electronic stopping loss of 430 MeV Kr irradiation, increasing temperature significantly enhances track visibility, enlarges track radii and promotes nanobubble formation. For higher electronic stopping conditions of 1171 MeV Ta irradiation, continuous ion tracks consisting of discontinuous nanobubbles (~1.5 nm radius) emerge already at 300 K, followed by a thermally activated transition into continuous channels with further radial expansion. At the atomic scale, SHI irradiation induces decomposition of wurtzite GaN into Ga clusters and N2 molecules along the ion trajectory, with Ga-rich regions and recrystallized wurtzite phases accumulating near bubble interfaces, while N2 preferentially segregates within bubble cores. Additionally, zincblende nanodomains nucleate around ion tracks and exhibit strong spatial correlation with radiation-induced dislocation networks, particularly screw dislocations, providing potential pathways for leakage current and increased susceptibility to single-event burnout.

cond-mat.mtrl-sci↗

arXiv2026

Reducing Detail Hallucinations in Long-Context Regulatory Understanding via Targeted Preference Optimization

Large language models (LLMs) frequently produce \emph{detail hallucinations} when processing long regulatory documents, including subtle errors in threshold values, units, scopes, obligation levels, and conditions that preserve surface plausibility while corrupting safety-critical parameters. We formalize this phenomenon through a fine-grained \emph{Detail Error Taxonomy} of five error types and introduce \textbf{DetailBench}, a benchmark built from 172 real regulatory documents and 150 synthetic documents spanning three jurisdictions, with human-annotated detail-level ground truth comprising 13,000 preference pairs. We propose \textbf{DetailDPO}, a targeted preference optimization framework that constructs contrastive pairs differing in exactly one detail dimension, concentrating DPO gradient signal on detail-bearing~tokens. We provide theoretical analysis showing why \emph{minimal detail perturbation} pairs yield gradient concentration under mild assumptions. Experiments on the Qwen2.5 family (7B, 14B, 72B) and Llama-3.1-8B across three context-length tiers (8K--64K tokens) show that DetailDPO reduces the Detail Error Rate by 42--61\% relative to baselines, with consistent gains across all five error types and cross-domain transfer to financial and medical documents.

cs.SI↗

arXiv2023

Coherent Dynamics of Charge Carriers in γ-InSe Revealed by Ultrafast Spectroscopy

For highly efficient ultrathin solar cells, layered indium selenide (InSe), a van der Waals solid, has shown a great promise. In this paper, we study the coherent dynamics of charge carriers generation in γ-InSe single crystals. We employ ultrafast transient absorption spectroscopy to examine the dynamics of hot electrons after resonant photoexcitation. To study the effect of excess kinetic energy of electrons after creating A exciton (VB1 to CB transition), we excite the sample with broadband pulses centered at 600, 650, 700 and 750 nm, respectively. We analyze the relaxation and recombination dynamics in γ-InSe by global fitting approach. Five decay associated spectra with their associated lifetimes are obtained, which have been assigned to intraband vibrational relaxation and interband recombination processes. We extract characteristic carrier thermalization times from 1 to 10 ps. To examine the coherent vibrations accompanying intraband relaxation dynamics, we analyze the kinetics by fitting to exponential functions and the obtained residuals are further processed for vibrational analysis. A few key phonon coherences are resolved and ab-initio quantum calculations reveal the nature of the associated phonons. The wavelet analysis is employed to study the time evolution of the observed coherences, which show that the low-frequency coherences last for more than 5 ps. Associated calculations reveal that the contribution of the intralayer phonon modes is the key determining factor for the scattering between free electrons and lattice. Our results provide fundamental insights into the photophysics in InSe and help to unravel their potential for high-performance optoelectronic devices.

cond-mat.mtrl-sci↗

arXiv2021

Multi-objective Feature Selection with Missing Data in Classification

Feature selection (FS) is an important research topic in machine learning. Usually, FS is modelled as a+ bi-objective optimization problem whose objectives are: 1) classification accuracy; 2) number of features. One of the main issues in real-world applications is missing data. Databases with missing data are likely to be unreliable. Thus, FS performed on a data set missing some data is also unreliable. In order to directly control this issue plaguing the field, we propose in this study a novel modelling of FS: we include reliability as the third objective of the problem. In order to address the modified problem, we propose the application of the non-dominated sorting genetic algorithm-III (NSGA-III). We selected six incomplete data sets from the University of California Irvine (UCI) machine learning repository. We used the mean imputation method to deal with the missing data. In the experiments, k-nearest neighbors (K-NN) is used as the classifier to evaluate the feature subsets. Experimental results show that the proposed three-objective model coupled with NSGA-III efficiently addresses the FS problem for the six data sets included in this study.

cs.AI↗

arXiv2020

Tuning the electronic properties of hydrogen passivated C3N nanoribbons through van der Waals stacking

The two-dimensional (2D) C3N has emerged as a material with promising applications in high performance device owing to its intrinsic bandgap and tunable electronic properties. Although there are several reports about the bandgap tuning of C3N via stacking or forming nanoribbon, bandgap modulation of bilayer C3N nanoribbons (C3NNRs) with various edge structures is still far from well understood. Here, based on extensive first-principles calculations, we demonstrated the effective bandgap engineering of C3N by cutting it into hydrogen passivated C3NNRs and stacking them into bilayer heterostructures. It was found that armchair (AC) C3NNRs with three types of edge structures are all semiconductors, while only zigzag (ZZ) C3NNRs with edges composed of both C and N atoms (ZZ-CN/CN) are semiconductors. The bandgaps of all semiconducting C3NNRs are larger than that of C3N nanosheet. More interestingly, AC-C3NNRs with CN/CN edges (AC-CN/CN) possess direct bandgap while ZZ-CN/CN have indirect bandgap. Compared with the monolayer C3NNR, the bandgaps of bilayer C3NNRs can be greatly modulated via different stacking orders and edge structures, varying from 0.43 eV for ZZ-CN/CN with AB'-stacking to 0.04 eV for AC-CN/CN with AA-stacking. Particularly, transition from direct to indirect bandgap was observed in the bilayer AC-CN/CN heterostructure with AA'-stacking, and the indirect-to-direct transition was found in the bilayer ZZ-CN/CN with AB-stacking. This work provides insights into the effective bandgap engineering of C3N and offers a new opportunity for its applications in nano-electronics and optoelectronic devices.

cond-mat.mtrl-sci↗