Search

With the rapid development of artificial intelligence, large language models (LLMs) have made remarkable advancements in natural language processing. These models are trained on vast datasets to exhibit powerful language understanding and generation capabilities across various applications, including machine translation, chatbots, and agents. However, LLMs have revealed a variety of privacy and security issues throughout their life cycle, drawing significant academic and industrial attention. Moreover, the risks faced by LLMs differ significantly from those encountered by traditional language models. Given that current surveys lack a clear taxonomy of unique threat models across diverse scenarios, we emphasize the unique privacy and security threats associated with five specific scenarios: pre-training, fine-tuning, retrieval-augmented generation systems, deployment, and LLM-based agents. Addressing the characteristics of each risk, this survey outlines potential threats and countermeasures. Research on attack and defense situations can offer feasible research directions, enabling more areas to benefit from LLMs.

Counterfactually Augmented Data (CAD) involves creating new data samples by applying minimal yet sufficient modifications to flip the label of existing data samples to other classes. Training with CAD enhances model robustness against spurious features that happen to correlate with labels by spreading the casual relationships across different classes. Yet, recent research reveals that training with CAD may lead models to overly focus on modified features while ignoring other important contextual information, inadvertently introducing biases that may impair performance on out-ofdistribution (OOD) datasets. To mitigate this issue, we employ contrastive learning to promote global feature alignment in addition to learning counterfactual clues. We theoretically prove that contrastive loss can encourage models to leverage a broader range of features beyond those modified ones. Comprehensive experiments on two human-edited CAD datasets demonstrate that our proposed method outperforms the state-of-the-art on OOD datasets., Comment: Accepted by ACL 2024 main conference

High-dimensional penalized rank regression is a powerful tool for modeling high-dimensional data due to its robustness and estimation efficiency. However, the non-smoothness of the rank loss brings great challenges to the computation. To solve this critical issue, high-dimensional convoluted rank regression is recently proposed, and penalized convoluted rank regression estimators are introduced. However, these developed estimators cannot be directly used to make inference. In this paper, we investigate the inference problem of high-dimensional convoluted rank regression. We first establish estimation error bounds of penalized convoluted rank regression estimators under weaker conditions on the predictors. Based on the penalized convoluted rank regression estimators, we further introduce a debiased estimator. We then provide Bahadur representation for our proposed estimator. We further develop simultaneous inference procedures. A novel bootstrap procedure is proposed and its theoretical validity is also established. Finally, simulation and real data analysis are conducted to illustrate the merits of our proposed methods.

In this paper, we construct and analyze new first- and second-order implicit-explicit (IMEX) schemes for the unsteady Navier-Stokes-Darcy model to describe the coupled free flow-porous media system, which is based on the scalar auxiliary variable (SAV) approach in time and finite element method in space. The constructed schemes are linear, only require solving a sequence of linear differential equations with constant coefficients at each time step, and can decouple the Navier-Stokes and Darcy systems. The unconditional stability of both the first- and second-order IMEX schemes can be derived for the coupled system equipped with the Lions interface condition, where the key point is that we should construct a new trilinear form to balance the fully explicit discretizations of the nonlinear terms in the complex system. We can also establish rigorous error estimates for the velocity and hydraulic head of the first-order scheme without any time step restriction. Numerical examples are presented to validate the proposed schemes.

Emerging unsupervised reconstruction techniques based on implicit neural representation (INR), such as NeRP, CoIL, and SCOPE, have shown unique capabilities in CT linear inverse imaging. In this work, we propose a novel unsupervised density neural representation (Diner) to tackle the challenging problem of CT metal artifacts when scanned objects contain metals. The drastic variation of linear attenuation coefficients (LACs) of metals over X-ray spectra leads to a nonlinear beam hardening effect (BHE) in CT measurements. Recovering CT images from metal-affected measurements therefore poses a complicated nonlinear inverse problem. Existing metal artifact reduction (MAR) techniques mostly formulate the MAR as an image inpainting task, which ignores the energy-induced BHE and produces suboptimal performance. Instead, our Diner introduces an energy-dependent polychromatic CT forward model to the INR framework, addressing the nonlinear nature of the MAR problem. Specifically, we decompose the energy-dependent LACs into energy-independent densities and energy-dependent mass attenuation coefficients (MACs) by fully considering the physical model of X-ray absorption. Using the densities as pivot variables and the MACs as known prior knowledge, the LACs can be accurately reconstructed from the raw measurements. Technically, we represent the unknown density map as an implicit function of coordinates. Combined with a novel differentiable forward model simulating the physical acquisition from the densities to the measurements, our Diner optimizes a multi-layer perception network to approximate the implicit function by minimizing predicted errors between the estimated and real measurements. Experimental results on simulated and real datasets confirm the superiority of our unsupervised Diner against popular supervised techniques in MAR performance and robustness., Comment: 13 pages

This paper is concerned with estimation and inference for ultrahigh dimensional partially linear single-index models. The presence of high dimensional nuisance parameter and nuisance unknown function makes the estimation and inference problem very challenging. In this paper, we first propose a profile partial penalized least squares estimator and establish the sparsity, consistency and asymptotic representation of the proposed estimator in ultrahigh dimensional setting. We then propose an $F$-type test statistic for parameters of primary interest and show that the limiting null distribution of the test statistic is $\chi^2$ distribution, and the test statistic can detect local alternatives, which converge to the null hypothesis at the root-$n$ rate. We further propose a new test for the specification testing problem of the nonparametric function. The test statistic is shown to be asymptotically normal. Simulation studies are conducted to examine the finite sample performance of the proposed estimators and tests. A real data example is used to illustrate the proposed procedures.

The surge in black-box AI models has prompted the need to explain the internal mechanism and justify their reliability, especially in high-stakes applications, such as healthcare and autonomous driving. Due to the lack of a rigorous definition of explainable AI (XAI), a plethora of research related to explainability, interpretability, and transparency has been developed to explain and analyze the model from various perspectives. Consequently, with an exhaustive list of papers, it becomes challenging to have a comprehensive overview of XAI research from all aspects. Considering the popularity of neural networks in AI research, we narrow our focus to a specific area of XAI research: gradient based explanations, which can be directly adopted for neural network models. In this review, we systematically explore gradient based explanation methods to date and introduce a novel taxonomy to categorize them into four distinct classes. Then, we present the essence of technique details in chronological order and underscore the evolution of algorithms. Next, we introduce both human and quantitative evaluations to measure algorithm performance. More importantly, we demonstrate the general challenges in XAI and specific challenges in gradient based explanations. We hope that this survey can help researchers understand state-of-the-art progress and their corresponding disadvantages, which could spark their interest in addressing these issues in future work.

The recent surge in research focused on generating synthetic data from large language models (LLMs), especially for scenarios with limited data availability, marks a notable shift in Generative Artificial Intelligence (AI). Their ability to perform comparably to real-world data positions this approach as a compelling solution to low-resource challenges. This paper delves into advanced technologies that leverage these gigantic LLMs for the generation of task-specific training data. We outline methodologies, evaluation techniques, and practical applications, discuss the current limitations, and suggest potential pathways for future research.

Stability is one of the key issues in mixed-halide perovskite alloys which are promising in emergent optoelectronics. Previous density-functional-theory (DFT) and machine learning studies indicate that the formation-energy convex hulls of these materials are very shallow, and stable alloy compositions are rare. In this work, we revisit this problem using DFT with special focus on the effects of configuration and vibration entropies. Allowed by the $20$-atomic models for the $\text{CsPb}(\text{I}_{1-x}^{}\text{Br}_x^{})_3^{}$ and $\text{CsPb}(\text{Br}_{1-x}^{}\text{Cl}_x^{})_3^{}$ series, the partition functions and therewith thermodynamic state functions are calculated by traversing all possible mixed-halide configurations. We can thus evaluate the temperature- and system-dependent configuration entropy, which largely corrects the conventional approach based on the ideal solution model. Finally, temperature-composition phase diagrams that include $\alpha$, $\beta$, $\gamma$ and $\delta$ phases of both alloys are constructed based on the free energy data, for which the contribution of phonon vibrations is included.

Utilizing synthetic dimensions generated by spatial or temporal modulation, topological pumping enables the exploration of higher-dimensional topological phenomena through lower-dimensional physical systems. In this letter, we propose a rational design paradigm of fast topological pumping based on 1D and 2D time-modulated discrete elastic lattices for the first time. Firstly, the realization of topological pumping is ensured by introducing quantitative indicators to drive a transition of the edge or corner state in the lattice spectrum. Meanwhile, with the help of limiting speed for adiabaticity to calculate the modulation time, a mathematical formulation of designing topological pumping with the fastest modulation speed is presented. By applying the proposed design paradigm, topological edge-bulk-edge and corner-bulk-corner energy transport are successfully achieved, with 11.2 and 4.0 times of improvement in modulation speed compared to classical pumping systems in the literature. In addition, applying to 1D and 2D space-modulated systems, the optimized modulation schemes can reduce the number of stacks to 5.3% and 26.8% of the classical systems while ensuring highly concentrated energy transport. This design paradigm is expected to be extended to the rational design of fast topological pumping in other physical fields.

Compared with traditional design methods, generative design significantly attracts engineers in various disciplines. In thiswork, howto achieve the real-time generative design of optimized structures with various diversities and controllable structural complexities is investigated. To this end, a modified Moving Morphable Component (MMC) method together with novel strategies are adopted to generate high-quality dataset. The complexity level of optimized structures is categorized by the topological invariant. By improving the cost function, the WGAN is trained to produce optimized designs with the input of loading position and complexity level in real time. It is found that, diverse designs with a clear load transmission path and crisp boundary, even not requiring further optimization and different from any reference in the dataset, can be generated by the proposed model. This method holds great potential for future applications of machine learning enhanced intelligent design.

The present paper introduces a hybrid explicit-implicit topology optimization method for shell-infill composite structure design. The proposed approach effectively combines the advantages of the explicit Moving Morphable Component (MMC) method, which describes structural topology only using a set of geometric parameters, and the implicit Solid Isotropic Material with Penalization (SIMP) method, which offers greater design freedom for characterizing the structural features. Compared to the existing methods for shell-infill structure design, the proposed approach can obtain optimized shell-infill structures with complex infill topology without resorting to complex filtering/projection operations. Numerical examples demonstrate the effectiveness of the proposed approach.

The task of multimodal relation extraction has attracted significant research attention, but progress is constrained by the scarcity of available training data. One natural thought is to extend existing datasets with cross-modal generative models. In this paper, we consider a novel problem setting, where only unimodal data, either text or image, are available during training. We aim to train a multimodal classifier from synthetic data that perform well on real multimodal test data. However, training with synthetic data suffers from two obstacles: lack of data diversity and label information loss. To alleviate the issues, we propose Mutual Information-aware Multimodal Iterated Relational dAta GEneration (MI2RAGE), which applies Chained Cross-modal Generation (CCG) to promote diversity in the generated data and exploits a teacher network to select valuable training samples with high mutual information with the ground-truth labels. Comparing our method to direct training on synthetic data, we observed a significant improvement of 24.06% F1 with synthetic text and 26.42% F1 with synthetic images. Notably, our best model trained on completely synthetic images outperforms prior state-of-the-art models trained on real multimodal data by a margin of 3.76% in F1. Our codebase will be made available upon acceptance.

Kagome lattice hosts a plethora of quantum states arising from the interplay of topology, spin-orbit coupling, and electron correlations. Here, we report symmetry-breaking electronic orders tunable by an applied magnetic field in a model Kagome magnet FeSn consisting of alternating stacks of two-dimensional Fe3Sn Kagome and Sn2 honeycomb layers. On the Fe3Sn layer terminated FeSn thin films epitaxially grown on SrTiO3(111) substrates, we observe trimerization of the Kagome lattice using scanning tunneling microscopy/spectroscopy, breaking its six-fold rotational symmetry while preserving the transitional symmetry. Such a trimerized Kagome lattice shows an energy-dependent contrast reversal in dI/dV maps, which is significantly enhanced by bound states induced by Sn vacancy defects. This trimerized Kagome lattice also exhibits stripe modulations that are energy-dependent and tunable by an applied in-plane magnetic field, indicating symmetry-breaking nematicity from the entangled magnetic and charge degrees of freedom in antiferromagnet FeSn.

Earth introduces strong attenuation and dispersion to propagating waves. The time-fractional wave equation with very small fractional exponent, based on Kjartansson's constant-Q theory, is widely recognized in the field of geophysics as a reliable model for frequency-independent Q anelastic behavior. Nonetheless, the numerical resolution of this equation poses considerable challenges due to the requirement of storing a complete time history of wavefields. To address this computational challenge, we present a novel approach: a nearly optimal sum-of-exponentials (SOE) approximation to the Caputo fractional derivative with very small fractional exponent, utilizing the machinery of generalized Gaussian quadrature. This method minimizes the number of memory variables needed to approximate the power attenuation law within a specified error tolerance. We establish a mathematical equivalence between this SOE approximation and the continuous fractional stress-strain relationship, relating it to the generalized Maxwell body model. Furthermore, we prove an improved SOE approximation error bound to thoroughly assess the ability of rheological models to replicate the power attenuation law. Numerical simulations on constant-Q viscoacoustic equation in 3D homogeneous media and variable-order P- and S- viscoelastic wave equations in 3D inhomogeneous media are performed. These simulations demonstrate that our proposed technique accurately captures changes in amplitude and phase resulting from material anelasticity. This advancement provides a significant step towards the practical usage of the time-fractional wave equation in seismic inversion., Comment: 28 pages, 52 figures, 6 tables

In this article, a novel explicit approach for designing complex thin-walled structures based on the Moving Morphable Component (MMC) method is proposed, which provides a unified framework to systematically address various design issues, including topology optimization, reinforced-rib layout optimization, and sandwich structure design problems. The complexity of thin-walled structures mainly comes from flexible geometries and the variation of thickness. On the one hand, the geometric complexity of thin-walled structures leads to the difficulty in automatically describing material distribution (e.g., reinforced ribs). On the other hand, thin-walled structures with different thicknesses require various hypotheses (e.g., Kirchhoff-Love shell theory and Reissner-Mindlin shell theory) to ensure the precision of structural responses. Whereas for cases that do not fit the shell hypothesis, the precision loss of response solutions is nonnegligible in the optimization process since the accumulation of errors will cause entirely different designs. Hence, the current article proposes a novel embedded solid component to tackle these challenges. The geometric constraints that make the components fit to the curved thin-walled structure are whereby satisfied. Compared with traditional strategies, the proposed method is free from the limit of shell assumptions of structural analysis and can achieve optimized designs with clear load transmission paths at the cost of few design variables and degrees of freedom for finite element analysis (FEA). Finally, we apply the proposed method to several representative examples to demonstrate its effectiveness, efficiency, versatility, and potential to handle complex industrial structures.

Contemporary news reporting increasingly features multimedia content, motivating research on multimedia event extraction. However, the task lacks annotated multimodal training data and artificially generated training data suffer from distribution shift from real-world data. In this paper, we propose Cross-modality Augmented Multimedia Event Learning (CAMEL), which successfully utilizes artificially generated multimodal training data and achieves state-of-the-art performance. We start with two labeled unimodal datasets in text and image respectively, and generate the missing modality using off-the-shelf image generators like Stable Diffusion and image captioners like BLIP. After that, we train the network on the resultant multimodal datasets. In order to learn robust features that are effective across domains, we devise an iterative and gradual training strategy. Substantial experiments show that CAMEL surpasses state-of-the-art (SOTA) baselines on the M2E2 benchmark. On multimedia events in particular, we outperform the prior SOTA by 4.2% F1 on event mention identification and by 9.8% F1 on argument identification, which indicates that CAMEL learns synergistic representations from the two modalities. Our work demonstrates a recipe to unleash the power of synthetic training data in structured prediction.

In real-world tasks, there is usually a large amount of unlabeled data and labeled data. The task of combining the two to learn is known as semi-supervised learning. Experts can use logical rules to label unlabeled data, but this operation is costly. The combination of perception and reasoning has a good effect in processing such semi-supervised tasks with domain knowledge. However, acquiring domain knowledge and the correction, reduction and generation of rules remain complex problems to be solved. Rough set theory is an important method for solving knowledge processing in information systems. In this paper, we propose a rule general abductive learning by rough set (RS-ABL). By transforming the target concept and sub-concepts of rules into information tables, rough set theory is used to solve the acquisition of domain knowledge and the correction, reduction and generation of rules at a lower cost. This framework can also generate more extensive negative rules to enhance the breadth of the knowledge base. Compared with the traditional semi-supervised learning method, RS-ABL has higher accuracy in dealing with semi-supervised tasks.

In this paper, we consider tests for ultrahigh-dimensional partially linear regression models. The presence of ultrahigh-dimensional nuisance covariates and unknown nuisance function makes the inference problem very challenging. We adopt machine learning methods to estimate the unknown nuisance function and introduce quadratic-form test statistics. Interestingly, though the machine learning methods can be very complex, under suitable conditions, we establish the asymptotic normality of our introduced test statistics under the null hypothesis and local alternative hypotheses. We further propose a power-enhanced procedure to improve the test statistics' performance. Two thresholding determination methods are provided for the power-enhanced procedure. We show that the power-enhanced procedure is powerful to detect signals under either sparse or dense alternatives and it can still control the type-I error asymptotically under the null hypothesis. Numerical studies are carried out to illustrate the empirical performance of our introduced procedures.

In this paper, we propose semiparametric efficient estimators of genetic relatedness between two traits in a model-free framework. Most existing methods require specifying certain parametric models involving the traits and genetic variants. However, the bias due to model misspecification may yield misleading statistical results. Moreover, the semiparametric efficient bounds for estimators of genetic relatedness are still lacking. In this paper, we develop semiparametric efficient estimators with machine learning methods and construct valid confidence intervals for two important measures of genetic relatedness: genetic covariance and genetic correlation, allowing both continuous and discrete responses. Based on the derived efficient influence functions of genetic relatedness, we propose a consistent estimator of the genetic covariance as long as one of genetic values is consistently estimated. The data of two traits may be collected from the same group or different groups of individuals. Various numerical studies are performed to illustrate our introduced procedures. We also apply proposed procedures to analyze Carworth Farms White mice genome-wide association study data., Comment: 46pages,9 tables, 1 figure

Structural disorder is common in metal-halide perovskites and important for understanding the functional properties of these materials. First-principles methods can address structure variation on the atomistic scale, but they are often limited by the lack of structure-sampling schemes required to characterize the disorder. In this work, structural disorder in the benchmark inorganic halide perovskites CsPbI$_3^{}$ and CsPbBr$_3^{}$ is computationally studied in terms of the three octahedral-tilting angles. The consequent variation in energetics and properties are described by three-dimensional potential-energy surfaces (PESs) and property landscapes, delivered by Bayesian Optimization Structure Search method with integrated density-functional-theory (DFT) calculations. The rapid convergence of the PES with about 200 DFT data points in three-dimensional searches demonstrates the power of active learning and strategic sampling with Bayesian optimization. Further analysis indicates that disorder grows with increasing temperature, and reveals that the materials band gap at finite temperatures is a statistical mean over disordered structures.

Numerical simulation of wave propagation in elastic media faces the challenges arising from increasing demand of high resolution in modern 3-D imaging applications, which requires a balance between efficiency and accuracy in addition to being friendly to the distributed high-performance computing environment. In this paper, we propose a distributed local spline simulator (LOSS) for solving the wave equation. LOSS uses patched cubic B-splines to represent the wavefields and attains an accurate evaluation of spatial derivatives with linear complexity. In order to link the adjacent patches, a perfectly matched boundary condition is introduced to give a closure of local spline coefficients. Owing to the rapid decay property of the local wavelets in dual space, it can recover the global spline as accurately as possible only at the cost of local communications among adjacent neighbors. Several typical numerical examples, including 2-D acoustic wave equation and P- and S- wave propagation in 3-D homogenous or heterogenous media, are provided to validate its convergence, accuracy and parallel scalability.