arXiv subjects

Guowei Yang

Publications and source records attributed to Guowei Yang.

At least 37 records · Page 2Linked to original sources

arXiv2025

XAMT: Cross-Framework API Matching for Testing Deep Learning Libraries

Deep learning powers critical applications such as autonomous driving, healthcare, and finance, where the correctness of underlying libraries is essential. Bugs in widely used deep learning APIs can propagate to downstream systems, causing serious consequences. While existing fuzzing techniques detect bugs through intra-framework testing across hardware backends (CPU vs. GPU), they may miss bugs that manifest identically across backends and thus escape detection under these strategies. To address this problem, we propose XAMT, a cross-framework fuzzing method that tests deep learning libraries by matching and comparing functionally equivalent APIs across different frameworks. XAMT matches APIs using similarity-based rules based on names, descriptions, and parameter structures. It then aligns inputs and applies variance-guided differential testing to detect bugs. We evaluated XAMT on five popular frameworks, including PyTorch, TensorFlow, Keras, Chainer, and JAX. XAMT matched 839 APIs and identified 238 matched API groups, and detected 17 bugs, 12 of which have been confirmed. Our results show that XAMT uncovers bugs undetectable by intra-framework testing, especially those that manifest consistently across backends. XAMT offers a complementary approach to existing methods and offers a new perspective on the testing of deep learning libraries.

cs.SE↗

arXiv2025

Harnessing LLMs for Document-Guided Fuzzing of OpenCV Library

The combination of computer vision and artificial intelligence is fundamentally transforming a broad spectrum of industries by enabling machines to interpret and act upon visual data with high levels of accuracy. As the biggest and by far the most popular open-source computer vision library, OpenCV library provides an extensive suite of programming functions supporting real-time computer vision. Bugs in the OpenCV library can affect the downstream computer vision applications, and it is critical to ensure the reliability of the OpenCV library. This paper introduces VISTAFUZZ, a novel technique for harnessing large language models (LLMs) for document-guided fuzzing of the OpenCV library. VISTAFUZZ utilizes LLMs to parse API documentation and obtain standardized API information. Based on this standardized information, VISTAFUZZ extracts constraints on individual input parameters and dependencies between these. Using these constraints and dependencies, VISTAFUZZ then generates new input values to systematically test each target API. We evaluate the effectiveness of VISTAFUZZ in testing 330 APIs in the OpenCV library, and the results show that VISTAFUZZ detected 17 new bugs, where 10 bugs have been confirmed, and 5 of these have been fixed.

cs.SE↗

arXiv2025

Enhancing LLM Code Generation with Ensembles: A Similarity-Based Selection Approach

Ensemble learning has been widely used in machine learning to improve model robustness, accuracy, and generalization, but has not yet been applied to code generation tasks with large language models (LLMs). We propose an ensemble approach for LLMs in code generation. Instead of relying on the output of a single model, we generate multiple candidate programs from different LLMs and apply a structured voting mechanism to select the most reliable solution. For voting, we compute syntactic and semantic similarity using CodeBLEU and behavioral equivalence using CrossHair's differential behavior analysis. By aggregating these similarity scores, we select the program that best aligns with the consensus among the candidates. We show through experiments that our ensemble approach consistently outperforms standalone LLMs on the well-known HumanEval and the more challenging LiveCodeBench datasets, achieving an accuracy of 90.2% and 50.2%, respectively, on the two datasets. In comparison, the best-performing LLM (GPT-4o) has an accuracy of 83.5% and 43.4%, respectively. Furthermore, even when restricted to free open-source models, our method achieves an accuracy of 80.5% and 41.6%, respectively, demonstrating the viability of our approach in resource-constrained settings.

cs.SE↗

arXiv2025

Large ferromagnetic-like band splitting in ultrathin ${\mathrm{SmC}}_{6}$ films

Two-dimensional (2D) magnetic materials provide a unique platform for exploring quantum phases from magnetic order in reduced dimensions. While there have been extensive studies on 2D magnetic materials based on 3$d$ electrons, experimental studies on 4$f$-electron counterparts are far fewer, particularly on their electronic structure. In this study, we report the successful synthesis of ultrathin ${\mathrm{SmC}}_{6}$ films using molecular beam epitaxy. Utilizing in situ angle-resolved photoemission spectroscopy (ARPES), we uncover a large band splitting in the valence bands, which we attribute to the ferromagnetic order driven by exchange couplings between Sm 4$f$ moments and conduction electrons. Despite the small magnetic moment of Sm, the observed splitting is comparable to those of Eu- and Gd-based systems with much larger local moments. Interestingly, the surface state also exhibits splitting with similar magnitude and can be eliminated by overannealing, while the valence bands with ferromagnetic-like splittings remain robust. Our work provides spectroscopic insight to understand the electronic origin of magnetic order in Sm-based compounds. Our study also offers a platform to study 2D magnetic materials based on 4$f$ electrons.

cond-mat.str-el↗

arXiv2025

$3d$ flat bands and coupled $4f$ moments in the kagome-honeycomb permanent magnet Sm$_{2}$Co$_{17}$

Rare earth permanent magnets (REPMs) with both localized moments and itinerant conduction bands are not only important for fundamental research but also have significant technological applications. In particular, Sm$_{\rm 2}$Co$_{\rm 17}$ is a prototypical high-temperture REPM, where the Co atoms form a kagome-honeycomb stacked lattice. Here we report synthesis of epitaxial Sm$_{\rm 2}$Co$_{\rm 17}$ films using molecular beam epitaxy and measurements of their momentum-resolved electronic structure from \textit{in-situ} angle-resolved photoemission spectroscopy. Our results unveil two flat bands from Co $3d$ orbitals near the Fermi level ($E_F$), one at $\sim$\,--300\,meV and another right at $E_F$, which arise from orbital-selective destructive interference and strong electron correlations, respectively. In addition, our results reveal that Sm $4f$ states are far away from $E_F$ (hence mostly localized) and exhibit an anomalous temperature dependence, caused by the $3d$-$4f$ magnetic coupling. Our findings provide direct spectroscopic insights to understand the strong uniaxial ferromagnetism in Sm$_{\rm 2}$Co$_{\rm 17}$ (and REPMs alike). Our work also opens avenues to explore flat-band physics near $E_F$ and emergent phenomena in correlated kagome-honeycomb lattices.

cond-mat.str-el↗

arXiv2025

Generating Realistic, Diverse, and Fault-Revealing Inputs with Latent Space Interpolation for Testing Deep Neural Networks

Deep Neural Networks (DNNs) have been widely employed across various domains, including safety-critical systems, necessitating comprehensive testing to ensure their reliability. Although numerous DNN model testing methods have been proposed to generate adversarial samples that are capable of revealing faults, existing methods typically perturb samples in the input space and then mutate these based on feedback from the DNN model. These methods often result in test samples that are not realistic and with low-probability reveal faults. To address these limitations, we propose a black-box DNN test input generation method, ARGUS, to generate realistic, diverse, and fault-revealing test inputs. ARGUS first compresses samples into a continuous latent space and then perturbs the original samples by interpolating these with samples of different classes. Subsequently, we employ a vector quantizer and decoder to reconstruct adversarial samples back into the input space. Additionally, we employ discriminators both in the latent space and in the input space to ensure the realism of the generated samples. Evaluation of ARGUS in comparison with state-of-the-art black-box testing and white-box testing methods, shows that ARGUS excels in generating realistic and diverse adversarial samples relative to the target dataset, and ARGUS successfully perturbs all original samples and achieves up to 4 times higher error rate than the best baseline method. Furthermore, using these adversarial samples for model retraining can improve model classification accuracy.

cs.LG↗

arXiv2025

Orientation-dependent superconductivity and electronic structure of the rare-earth metal/KTaO3 interfaces

The recent discovery of orientation-dependent superconductivity in KTaO3-based interfaces has attracted considerable interest, while the underlying origin remains an open question. Here we report a different approach to tune the interfacial electron gas and superconductivity by forming interfaces between rare-earth (RE) metals (RE being La, Ce, Eu) and KTaO3 substrates with different orientations. We found that the interfacial superconductivity is strongest for the Eu/KTaO3 interfaces, becomes weaker in La/KTaO3 and is absent in Ce/KTaO3. Using in-situ photoemission, we observed distinct valence bands associated with RE metals, as well as a pronounced orientation dependence in the interfacial electronic structure, which can be linked to the orientation-dependent superconductivity. The photoemission spectra show similar double-peak structures for the (111) and (110) oriented interfaces, with an energy separation close to the LO4 phonon of KTaO3. Detailed analyses suggest that this double-peak structure could be attributed to electron-phonon coupling, which might be important for the interfacial superconductivity.

cond-mat.supr-con↗

arXiv2025

Mimicking the Familiar: Dynamic Command Generation for Information Theft Attacks in LLM Tool-Learning System

Information theft attacks pose a significant risk to Large Language Model (LLM) tool-learning systems. Adversaries can inject malicious commands through compromised tools, manipulating LLMs to send sensitive information to these tools, which leads to potential privacy breaches. However, existing attack approaches are black-box oriented and rely on static commands that cannot adapt flexibly to the changes in user queries and the invocation chain of tools. It makes malicious commands more likely to be detected by LLM and leads to attack failure. In this paper, we propose AutoCMD, a dynamic attack comment generation approach for information theft attacks in LLM tool-learning systems. Inspired by the concept of mimicking the familiar, AutoCMD is capable of inferring the information utilized by upstream tools in the toolchain through learning on open-source systems and reinforcement with target system examples, thereby generating more targeted commands for information theft. The evaluation results show that AutoCMD outperforms the baselines with +13.2% $ASR_{Theft}$, and can be generalized to new tool-learning systems to expose their information leakage risks. We also design four defense methods to effectively protect tool-learning systems from the attack.

cs.AI↗

arXiv2025

Three-dimensional mapping of the altermagnetic spin splitting in CrSb

Altermagnetism, a kind of collinear magnetism that is characterized by a momentum-dependent band and spin splitting without net magnetization, has recently attracted considerable interest. Finding altermagnetic materials with large splitting near the Fermi level necessarily requires three-dimensional k-space mapping. While this is crucial for spintronic applications and emergent phenomena, it remains challenging. Here, using synchrotron-based angle-resolved photoemission spectroscopy (ARPES), spin-resolved ARPES and model calculations, we uncover a large altermagnetic splitting, up to ~1.0 eV, near the Fermi level in CrSb. We verify its bulk-type g-wave altermagnetism through systematic three-dimensional k-space mapping, which unambiguously reveals the altermagnetic symmetry and associated nodal planes. Spin-resolved ARPES measurements further verify the spin polarizations of the split bands near Fermi level. Tight-binding model analysis indicates that the large altermagnetic splitting arises from strong third-nearest-neighbor hopping mediated by Sb ions. The large band/spin splitting near Fermi level in metallic CrSb, together with its high TN (up to 705 K) and simple spin configuration, paves the way for exploring emergent phenomena and spintronic applications based on altermagnets.

cond-mat.mtrl-sci↗

arXiv2024

A Universal Method to Transform Aromatic Hydrocarbon Molecules into Confined Carbyne inside Single-Walled Carbon Nanotubes

Carbyne, a sp1-hybridized allotrope of carbon, is a linear carbon chain with exceptional theoretically predicted properties that surpass those of sp2-hybridized graphene and carbon nanotubes (CNTs). However, the existence of carbyne has been debated due to its instability caused by Peierls distortion, which limits its practical development. The only successful synthesis of carbyne has been achieved inside CNTs, resulting in a form known as confined carbyne (CC). However, CC can only be synthesized inside multi-walled CNTs, limiting its property-tuning capabilities to the inner tubes of the CNTs. Here, we present a universal method for synthesizing CC inside single-walled carbon nanotubes (SWCNTs) with diameter of 0.9-1.3 nm. Aromatic hydrocarbon molecules are filled inside SWCNTs and subsequently transformed into CC under low-temperature annealing. A variety of aromatic hydrocarbon molecules are confirmed as effective precursors for formation of CC, with Raman frequencies centered around 1861 cm-1. Enriched (6,5) and (7,6) SWCNTs with diameters less than 0.8 nm are less effective than the SWCNTs with diameter of 0.9-1.3 nm for CC formation. Furthermore, resonance Raman spectroscopy reveals that optical band gap of the CC at 1861 cm-1 is 2.353 eV, which is consistent with the result obtained using a linear relationship between the Raman signal and optical band gap. This newly developed approach provides a versatile route for synthesizing CC from various precursor molecules inside diverse templates, which is not limited to SWCNTs but could extend to any templates with appropriate size, including molecular sieves, zeolites, boron nitride nanotubes, and metal-organic frameworks.

cond-mat.mtrl-sci↗

arXiv2024

Automated Update of Android Deprecated API Usages with Large Language Models

Android apps rely on application programming interfaces (APIs) to access various functionalities of Android devices. These APIs however are regularly updated to incorporate new features while the old APIs get deprecated. Even though the importance of updating deprecated API usages with the recommended replacement APIs has been widely recognized, it is non-trivial to update the deprecated API usages. Therefore, the usages of deprecated APIs linger in Android apps and cause compatibility issues in practice. This paper introduces GUPPY, an automated approach that utilizes large language models (LLMs) to update Android deprecated API usages. By employing carefully crafted prompts, GUPPY leverages GPT-4, one of the most powerful LLMs, to update deprecated-API usages, ensuring compatibility in both the old and new API levels. Additionally, GUPPY uses GPT-4 to generate tests, identify incorrect updates, and refine the API usage through an iterative process until the tests pass or a specified limit is reached. Our evaluation, conducted on 360 benchmark API usages from 20 deprecated APIs and an additional 156 deprecated API usages from the latest API levels 33 and 34, demonstrates GUPPY's advantages over the state-of-the-art techniques.

cs.SE↗

arXiv2024

PatUntrack: Automated Generating Patch Examples for Issue Reports without Tracked Insecure Code

Security patches are essential for enhancing the stability and robustness of projects in the software community. While vulnerabilities are officially expected to be patched before being disclosed, patching vulnerabilities is complicated and remains a struggle for many organizations. To patch vulnerabilities, security practitioners typically track vulnerable issue reports (IRs), and analyze their relevant insecure code to generate potential patches. However, the relevant insecure code may not be explicitly specified and practitioners cannot track the insecure code in the repositories, thus limiting their ability to generate patches. In such cases, providing examples of insecure code and the corresponding patches would benefit the security developers to better locate and fix the insecure code. In this paper, we propose PatUntrack to automatically generating patch examples from IRs without tracked insecure code. It auto-prompts Large Language Models (LLMs) to make them applicable to analyze the vulnerabilities. It first generates the completed description of the Vulnerability-Triggering Path (VTP) from vulnerable IRs. Then, it corrects hallucinations in the VTP description with external golden knowledge. Finally, it generates Top-K pairs of Insecure Code and Patch Example based on the corrected VTP description. To evaluate the performance, we conducted experiments on 5,465 vulnerable IRs. The experimental results show that PatUntrack can obtain the highest performance and improve the traditional LLM baselines by +14.6% (Fix@10) on average in patch example generation. Furthermore, PatUntrack was applied to generate patch examples for 76 newly disclosed vulnerable IRs. 27 out of 37 replies from the authors of these IRs confirmed the usefulness of the patch examples generated by PatUntrack, indicating that they can benefit from these examples for patching the vulnerabilities.

cs.CR↗

arXiv2024

MTDSense: AI-Based Fingerprinting of Moving Target Defense Techniques in Software-Defined Networking

Moving target defenses (MTD) are proactive security techniques that enhance network security by confusing the attacker and limiting their attack window. MTDs have been shown to have significant benefits when evaluated against traditional network attacks, most of which are automated and untargeted. However, little has been done to address an attacker who is aware the network uses an MTD. In this work, we propose a novel approach named MTDSense, which can determine when the MTD has been triggered using the footprints the MTD operation leaves in the network traffic. MTDSense uses unsupervised clustering to identify traffic following an MTD trigger and extract the MTD interval. An attacker can use this information to maximize their attack window and tailor their attacks, which has been shown to significantly reduce the effectiveness of MTD. Through analyzing the attacker's approach, we propose and evaluate two new MTD update algorithms that aim to reduce the information leaked into the network by the MTD. We present an extensive experimental evaluation by creating, to our knowledge, the first dataset of the operation of an IP-shuffling MTD in a software-defined network. Our work reveals that despite previous results showing the effectiveness of MTD as a defense, traditional implementations of MTD are highly susceptible to a targeted attacker.

cs.CR↗

arXiv2024

On the Preprocessing of Physics-informed Neural Networks: How to Better Utilize Data in Fluid Mechanics

Physics-Informed Neural Networks (PINNs) serve as a flexible alternative for tackling forward and inverse problems in differential equations, displaying impressive advancements in diverse areas of applied mathematics. Despite integrating both data and underlying physics to enrich the neural network's understanding, concerns regarding the effectiveness and practicality of PINNs persist. Over the past few years, extensive efforts in the current literature have been made to enhance this evolving method, by drawing inspiration from both machine learning algorithms and numerical methods. Despite notable progressions in PINNs algorithms, the important and fundamental field of data preprocessing remain unexplored, limiting the applications of PINNs especially in solving inverse problems. Therefore in this paper, a concise yet potent data preprocessing method focusing on data normalization was proposed. By applying a linear transformation to both the data and corresponding equations concurrently, the normalized PINNs approach was evaluated on the task of reconstructing flow fields in three turbulent cases. The results illustrate that by adhering to the data preprocessing procedure, PINNs can robustly achieve higher prediction accuracy for all flow quantities under different hyperparameter setups, without incurring extra computational cost, distinctly improving the utilization of limited training data. Though only verified in Navier-Stokes (NS) equations, this method holds potential for application to various other equations.

physics.flu-dyn↗

arXiv2024

Fermi Surface Nesting with Heavy Quasiparticles in the Locally Noncentrosymmetric Superconductor CeRh$_2$As$_2$

The locally noncentrosymmetric heavy fermion superconductor CeRh$_2$As$_2$ has attracted considerable interests due to its rich superconducting phases, accompanied by a quadrupole density wave and pronounced antiferromagnetic excitations. To understand the underlying physics, we here report measurements from high-resolution angle-resolved photoemission. Our results reveal fine splittings of the conduction bands related to the locally noncentrosymmetric structure, as well as a quasi-two-dimensional Fermi surface (FS) with strong $4f$ contributions. The FS exhibits nesting with an in-plane vector $(π/a, π/a)$, which is facilitated by the van Hove singularity near $\bar X$ that arises from the characteristic conduction-$f$ hybridization. The FS nesting provides a natural explanation for the observed antiferromagnetic excitations at $(π/a, π/a)$, which could be intimately connected to its unconventional superconductivity. Our experimental results are well supported by density functional theory plus dynamical mean field theory calculations, which can capture the strong correlation effects. Our study not only provides spectroscopic proof of the key factors underlying the field-induced superconducting transition, but also uncovers the critical role of FS nesting and lattice Kondo effect in the intertwined spin and charge fluctuations.

cond-mat.str-el↗

arXiv2024

Mirage: An RNS-Based Photonic Accelerator for DNN Training

Photonic computing is a compelling avenue for performing highly efficient matrix multiplication, a crucial operation in Deep Neural Networks (DNNs). While this method has shown great success in DNN inference, meeting the high precision demands of DNN training proves challenging due to the precision limitations imposed by costly data converters and the analog noise inherent in photonic hardware. This paper proposes Mirage, a photonic DNN training accelerator that overcomes the precision challenges in photonic hardware using the Residue Number System (RNS). RNS is a numeral system based on modular arithmetic, allowing us to perform high-precision operations via multiple low-precision modular operations. In this work, we present a novel micro-architecture and dataflow for an RNS-based photonic tensor core performing modular arithmetic in the analog domain. By combining RNS and photonics, Mirage provides high energy efficiency without compromising precision and can successfully train state-of-the-art DNNs achieving accuracy comparable to FP32 training. Our study shows that on average across several DNNs when compared to systolic arrays, Mirage achieves more than $23.8\times$ faster training and $32.1\times$ lower EDP in an iso-energy scenario and consumes $42.8\times$ lower power with comparable or better EDP in an iso-area scenario.

cs.AR↗

arXiv2024

A Framework of Data Assimilation for Wind Flow Fields by Physics-informed Neural Networks

Various types of measurement techniques, such as Light Detection and Ranging (LiDAR) devices, anemometers, and wind vanes, are extensively utilized in wind energy to characterize the inflow. However, these methods typically gather data at limited points within local wind fields, capturing only a fraction of the wind field's characteristics at wind turbine sites, thus hindering detailed wind field analysis. This study introduces a framework using Physics-informed Neural Networks to assimilate diverse sensor data types. This includes line-of-sight wind speed, velocity magnitude and direction, velocity components, and pressure. Moreover, the parameterized Navier-Stokes equations are integrated as physical constraints, ensuring that the neural networks accurately represent atmospheric flow dynamics. The framework accounts for the turbulent nature of atmospheric boundary layer flow by including artificial eddy viscosity in the network outputs, enhancing the model's ability to learn and accurately depict large-scale flow structures. The reconstructed flow field and the effective wind speed are in good agreement with the actual data. Furthermore, a transfer learning strategy is employed for the online deployment of pre-trained PINN, which requires less time than that of the actual physical flow. This capability allows the framework to reconstruct wind flow fields in real time based on live data. In the demo cases, the maximum error between the effective wind speed reconstructed online and the actual value at the wind turbine site is only 3.7%. The proposed data assimilation framework provides a universal tool for reconstructing spatiotemporal wind flow fields using various measurement data. Additionally, it presents a viable approach for the online assimilation of real-time measurements. To facilitate the utilization of wind energy, our framework's source code is openly accessible.

physics.flu-dyn↗

arXiv2024

Groot: Adversarial Testing for Generative Text-to-Image Models with Tree-based Semantic Transformation

With the prevalence of text-to-image generative models, their safety becomes a critical concern. adversarial testing techniques have been developed to probe whether such models can be prompted to produce Not-Safe-For-Work (NSFW) content. However, existing solutions face several challenges, including low success rate and inefficiency. We introduce Groot, the first automated framework leveraging tree-based semantic transformation for adversarial testing of text-to-image models. Groot employs semantic decomposition and sensitive element drowning strategies in conjunction with LLMs to systematically refine adversarial prompts. Our comprehensive evaluation confirms the efficacy of Groot, which not only exceeds the performance of current state-of-the-art approaches but also achieves a remarkable success rate (93.66%) on leading text-to-image models such as DALL-E 3 and Midjourney.

cs.CL↗