arXiv subjects

Xi Tang

Publications and source records attributed to Xi Tang.

14 recordsLinked to original sources

arXiv2025

Semantically-Equivalent Transformations-Based Backdoor Attacks against Neural Code Models: Characterization and Mitigation

Neural code models have been increasingly incorporated into software development processes. However, their susceptibility to backdoor attacks presents a significant security risk. The state-of-the-art understanding focuses on injection-based attacks, which insert anomalous patterns into software code. These attacks can be neutralized by standard sanitization techniques. This status quo may lead to a false sense of security regarding backdoor attacks. In this paper, we introduce a new kind of backdoor attacks, dubbed Semantically-Equivalent Transformation (SET)-based backdoor attacks, which use semantics-preserving low-prevalence code transformations to generate stealthy triggers. We propose a framework to guide the generation of such triggers. Our experiments across five tasks, six languages, and models like CodeBERT, CodeT5, and StarCoder show that SET-based attacks achieve high success rates (often >90%) while preserving model utility. The attack proves highly stealthy, evading state-of-the-art defenses with detection rates on average over 25.13% lower than injection-based counterparts. We evaluate normalization-based countermeasures and find they offer only partial mitigation, confirming the attack's robustness. These results motivate further investigation into scalable defenses tailored to SET-based attacks.

cs.SE↗

arXiv2025

Modeling of Parallel Single-Pixel Imaging for 3D Reconstruction: New Insights and Opportunities

The growing prevalence of intelligent manufacturing and autonomous vehicles has intensified the demand for three-dimensional (3D) reconstruction under complex reflection and transmission conditions. Traditional structured light techniques rely on inherent point-to-point triangulation, which limits accurate 3D measurements in these challenging scenarios. Parallel single-pixel imaging (PSI) has demonstrated unprecedented superiority under extreme conditions and has emerged as a promising approach of accurate 3D measurements. However, a complete theoretical model has not been reported in existing work to well explain its underlying mechanisms and quantitatively characterize its performance. In this study, a comprehensive theoretical model for the PSI method is proposed, including imaging and noise models. The proposed imaging model describes light transport coefficients under complex illumination, elucidating the intrinsic mechanisms of successful 3D imaging using PSI. The developed noise model quantitatively analyzes the impact of environmental noise on measurement accuracy, offering a framework to guide the error analysis of a PSI system. Numerical simulations and experimental results validate the proposed models, revealing the generality and robustness of PSI. Finally, potential research directions are highlighted to guide and inspire future investigations. The established theoretical models lay a solid foundation of PSI and brings new insights and opportunities for future application in more demanding 3D reconstruction tasks.

physics.optics↗

arXiv2025

Acyclicity test of complexes modulo Serre subcategories using the residue fields

Let $R$ be a commutative noetherian ring, and let $\mathscr{S}$(resp. $\mathscr{L}$) be a Serre(resp. localizing) subcategory of the category of $R$-modules. If $\Bbb F$ is an unbounded complex of $R$-modules Tor-perpendicular to $\mathscr{S}$ and $d$ is an integer, then $\HH{i\geqslant d}{S\otimes_R \Bbb F}$ is in $\mathscr{L}$ for each $R$-module $S$ in $\mathscr{S}$ if and only if $\HH{i\geqslant d}{k(\fp)\otimes_R \Bbb F}$ is in $\mathscr{L}$ for each prime ideal $\fp$ such that $R/\fp$ is in $\mathscr{S}$, where $k(\fp)$ is the residue field at $\fp$. As an application, we show that for any $R$-module $M$, $\Tor_{i\geqslant 0}^R(k(\fp),M)$ is in $\mathscr{L}$ for each prime ideal $\fp$ such that $R/\fp$ is in $\mathscr{S}$ if and only if $\Ext^{i \geqslant 0}_R(S,M)$ is in $\mathscr{L}$ for each cyclic $R$-module $S$ in $\mathscr{S}$. We also obtain some new characterizations of regular and Gorenstein rings in the case of $\mathscr{S}$ consists of finite modules with supports in a specialization-closed subset $V(I)$ of $\Spec R$.

math.AC↗

arXiv2025

Adaptive Keyframe Sampling for Long Video Understanding

Multimodal large language models (MLLMs) have enabled open-world visual understanding by injecting visual input as extra tokens into large language models (LLMs) as contexts. However, when the visual input changes from a single image to a long video, the above paradigm encounters difficulty because the vast amount of video tokens has significantly exceeded the maximal capacity of MLLMs. Therefore, existing video-based MLLMs are mostly established upon sampling a small portion of tokens from input data, which can cause key information to be lost and thus produce incorrect answers. This paper presents a simple yet effective algorithm named Adaptive Keyframe Sampling (AKS). It inserts a plug-and-play module known as keyframe selection, which aims to maximize the useful information with a fixed number of video tokens. We formulate keyframe selection as an optimization involving (1) the relevance between the keyframes and the prompt, and (2) the coverage of the keyframes over the video, and present an adaptive algorithm to approximate the best solution. Experiments on two long video understanding benchmarks validate that Adaptive Keyframe Sampling improves video QA accuracy (beyond strong baselines) upon selecting informative keyframes. Our study reveals the importance of information pre-filtering in video-based MLLMs. Code is available at https://github.com/ncTimTang/AKS.

cs.CV↗

arXiv2025

IMDPrompter: Adapting SAM to Image Manipulation Detection by Cross-View Automated Prompt Learning

Using extensive training data from SA-1B, the Segment Anything Model (SAM) has demonstrated exceptional generalization and zero-shot capabilities, attracting widespread attention in areas such as medical image segmentation and remote sensing image segmentation. However, its performance in the field of image manipulation detection remains largely unexplored and unconfirmed. There are two main challenges in applying SAM to image manipulation detection: a) reliance on manual prompts, and b) the difficulty of single-view information in supporting cross-dataset generalization. To address these challenges, we develops a cross-view prompt learning paradigm called IMDPrompter based on SAM. Benefiting from the design of automated prompts, IMDPrompter no longer relies on manual guidance, enabling automated detection and localization. Additionally, we propose components such as Cross-view Feature Perception, Optimal Prompt Selection, and Cross-View Prompt Consistency, which facilitate cross-view perceptual learning and guide SAM to generate accurate masks. Extensive experimental results from five datasets (CASIA, Columbia, Coverage, IMD2020, and NIST16) validate the effectiveness of our proposed method.

cs.CV↗

arXiv2025

Rethinking Pseudo-Label Guided Learning for Weakly Supervised Temporal Action Localization from the Perspective of Noise Correction

Pseudo-label learning methods have been widely applied in weakly-supervised temporal action localization. Existing works directly utilize weakly-supervised base model to generate instance-level pseudo-labels for training the fully-supervised detection head. We argue that the noise in pseudo-labels would interfere with the learning of fully-supervised detection head, leading to significant performance leakage. Issues with noisy labels include:(1) inaccurate boundary localization; (2) undetected short action clips; (3) multiple adjacent segments incorrectly detected as one segment. To target these issues, we introduce a two-stage noisy label learning strategy to harness every potential useful signal in noisy labels. First, we propose a frame-level pseudo-label generation model with a context-aware denoising algorithm to refine the boundaries. Second, we introduce an online-revised teacher-student framework with a missing instance compensation module and an ambiguous instance correction module to solve the short-action-missing and many-to-one problems. Besides, we apply a high-quality pseudo-label mining loss in our online-revised teacher-student framework to add different weights to the noisy labels to train more effectively. Our model outperforms the previous state-of-the-art method in detection accuracy and inference speed greatly upon the THUMOS14 and ActivityNet v1.2 benchmarks.

cs.CV↗

arXiv2024

Weakly Supervised Temporal Action Localization via Dual-Prior Collaborative Learning Guided by Multimodal Large Language Models

Recent breakthroughs in Multimodal Large Language Models (MLLMs) have gained significant recognition within the deep learning community, where the fusion of the Video Foundation Models (VFMs) and Large Language Models(LLMs) has proven instrumental in constructing robust video understanding systems, effectively surmounting constraints associated with predefined visual tasks. These sophisticated MLLMs exhibit remarkable proficiency in comprehending videos, swiftly attaining unprecedented performance levels across diverse benchmarks. However, their operation demands substantial memory and computational resources, underscoring the continued importance of traditional models in video comprehension tasks. In this paper, we introduce a novel learning paradigm termed MLLM4WTAL. This paradigm harnesses the potential of MLLM to offer temporal action key semantics and complete semantic priors for conventional Weakly-supervised Temporal Action Localization (WTAL) methods. MLLM4WTAL facilitates the enhancement of WTAL by leveraging MLLM guidance. It achieves this by integrating two distinct modules: Key Semantic Matching (KSM) and Complete Semantic Reconstruction (CSR). These modules work in tandem to effectively address prevalent issues like incomplete and over-complete outcomes common in WTAL methods. Rigorous experiments are conducted to validate the efficacy of our proposed approach in augmenting the performance of various heterogeneous WTAL models.

cs.CV↗

arXiv2024

Artemis: Towards Referential Understanding in Complex Videos

Videos carry rich visual information including object description, action, interaction, etc., but the existing multimodal large language models (MLLMs) fell short in referential understanding scenarios such as video-based referring. In this paper, we present Artemis, an MLLM that pushes video-based referential understanding to a finer level. Given a video, Artemis receives a natural-language question with a bounding box in any video frame and describes the referred target in the entire video. The key to achieving this goal lies in extracting compact, target-specific video features, where we set a solid baseline by tracking and selecting spatiotemporal features from the video. We train Artemis on the newly established VideoRef45K dataset with 45K video-QA pairs and design a computationally efficient, three-stage training procedure. Results are promising both quantitatively and qualitatively. Additionally, we show that \model can be integrated with video grounding and text summarization tools to understand more complex scenarios. Code and data are available at https://github.com/qiujihao19/Artemis.

cs.CV↗

arXiv2024

ChatterBox: Multi-round Multimodal Referring and Grounding

In this study, we establish a baseline for a new task named multimodal multi-round referring and grounding (MRG), opening up a promising direction for instance-level multimodal dialogues. We present a new benchmark and an efficient vision-language model for this purpose. The new benchmark, named CB-300K, spans challenges including multi-round dialogue, complex spatial relationships among multiple instances, and consistent reasoning, which are beyond those shown in existing benchmarks. The proposed model, named ChatterBox, utilizes a two-branch architecture to collaboratively handle vision and language tasks. By tokenizing instance regions, the language branch acquires the ability to perceive referential information. Meanwhile, ChatterBox feeds a query embedding in the vision branch to a token receiver for visual grounding. A two-stage optimization strategy is devised, making use of both CB-300K and auxiliary external data to improve the model's stability and capacity for instance-level understanding. Experiments show that ChatterBox outperforms existing models in MRG both quantitatively and qualitatively, paving a new path towards multimodal dialogue scenarios with complicated and precise interactions. Code, data, and model are available at: https://github.com/sunsmarterjie/ChatterBox.

cs.CV↗

arXiv2019

Higher differential objects in additive categories

Given an additive category $\mathcal{C}$ and an integer $n\geqslant 2$. We form a new additive category $\mathcal{C}[\epsilon]^n$ consisting of objects $X$ in $\mathcal{C}$ equipped with an endomorphism $\epsilon_X$ satisfying ${\epsilon^n_X}=0$. First, using the descriptions of projective and injective objects in $\mathcal{C}[\epsilon]^n$, we not only establish a connection between Gorenstein flat modules over a ring $R$ and $R[t]/(t^n)$, but also prove that an Artinian algebra $R$ satisfies some homological conjectures if and only if so does $R[t]/(t^n)$. Then we show that the corresponding homotopy category $\K(\mathcal{C}[\epsilon]^n)$ is a triangulated category when $\mathcal{C}$ is an idempotent complete exact category. Moreover, under some conditions for an abelian category $\mathcal{A}$, the natural quotient functor $Q$ from $\K(\mathcal{A}[\epsilon]^n)$ to the derived category $\D(\mathcal{A}[\epsilon]^n)$ produces a recollement of triangulated categories. Finally, we prove that if $\mathcal{A}$ is an Ab4-category with a compact projective generator, then $\D(\mathcal{A}[\epsilon]^n)$ is a compactly generated triangulated category.

math.RT↗

arXiv2019

Cograde conditions and cotorsion pairs

Let $R$ and $S$ be rings and $_R\omega_S$ a semidualizing bimodule. We study when the double functor $\Tor^S_i(\omega, \Ext^i_{R}(\omega,-))$ preserves epimorphisms and the double functor $\Ext_{R}^i(\omega, \Tor_i^{S}(\omega,-))$ preserves monomorphisms in terms of the (strong) cograde conditions of modules. Under certain cograde condition of modules, we construct two complete cotorsion pairs. In addition, we establish the relation between some relative finitistic dimensions of rings and the right and left projective dimensions of $\omega$.

math.RA↗

arXiv2018

Barrier Inhomogeneity of Schottky Diode on Nonpolar AlN Grown by Physical Vapor Transport

An aluminum nitride (AlN) Schottky barrier diode (SBD) was fabricated on a nonpolar AlN crystal grown on tungsten substrate by physical vapor transport. The Ni/Au-AlN SBD features a low ideality factor n of 3.3 and an effective Schottky barrier height (SBH) of 1.05 eV at room temperature. The ideality factor n decreases and the effective SBH increases at high temperatures. The temperature dependences of n and SBH were explained using an inhomogeneous model. A mean SBH of 2.105 eV was obtained for the Ni-AlN Schottky junction from the inhomogeneity analysis of the current-voltage characteristics. An equation in which the parameters have explicit physical meanings in thermionic emission theory is proposed to describe the current-voltage characteristics of inhomogeneous SBDs.

physics.app-ph↗

arXiv2015

Homological Aspects of the Dual Auslander Transpose II

Let $R$ and $S$ be rings and $_R\omega_S$ a semidualizing bimodule. We prove that there exists a Morita equivalence between the class of $\infty$-$\omega$-cotorsionfree modules and a subclass of the class of $\omega$-adstatic modules. Also we establish the relation between the relative homological dimensions of a module $M$ and the corresponding standard homological dimensions of ${\rm Hom}(\omega,M)$. By investigating the properties of the Bass injective dimension of modules (resp. complexes), we get some equivalent characterizations of semi-tilting modules (resp. Gorenstein artin algebras). Finally we obtain a dual version of the Auslander-Bridger's approximation theorem. As a consequence, we get some equivalent characterizations of Auslander $n$-Gorenstein artin algebras.

math.RA↗

arXiv2015

Homological Aspects of the Dual Auslander Transpose

As a dual of the Auslander transpose of modules, we introduce and study the cotranspose of modules with respect to a semidualizing module $C$. Then using it we introduce $n$-$C$-cotorsionfree modules, and show that $n$-$C$-cotorsionfree modules possess many dual properties of $n$-torsionfree modules. In particular, we show that $n$-$C$-cotorsionfree modules are useful in characterizing the Bass class and investigating the approximation theory for modules. Moreover, we study $n$-cotorsionfree modules over artin algebras and answer negatively an open question of Huang and Huang posed in 2012.

math.KT↗