Author name cluster

Chengze Jiang

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

3 papers

1 author row

AAAI Conference 2026 Conference Paper

Diversifying Counterattacks: Orthogonal Exploration for Robust CLlP Inference

Chengze Jiang
Minjing Dong
Xinli Shi
Jie Gui

Vision-language pre-training models (VLPs) demonstrate strong multimodal understanding and zero-shot generalization, yet remain vulnerable to adversarial examples, raising concerns about their reliability. Recent work, Test-Time Counterattack (TTC), improves robustness by generating perturbations that maximize the embedding deviation of adversarial inputs using PGD, pushing them away from their adversarial representations. However, due to the fundamental difference in optimization objectives between adversarial attacks and counterattacks, generating counterattacks solely based on gradients with respect to the adversarial input confines the search to a narrow space. As a result, the counterattacks could overfit limited adversarial patterns and lack the diversity to fully neutralize a broad range of perturbations. In this work, we argue that enhancing the diversity and coverage of counterattacks is crucial to improving adversarial robustness in test-time defense. Accordingly, we propose Directional Orthogonal Counterattack (DOC), which augments counterattack optimization by incorporating orthogonal gradient directions and momentum-based updates. This design expands the exploration of the counterattack space and increases the diversity of perturbations, which facilitates the discovery of more generalizable counterattacks and ultimately improves the ability to neutralize adversarial perturbations. Meanwhile, we present a directional sensitivity score based on averaged cosine similarity to boost DOC by improving example discrimination and adaptively modulating the counterattack strength. Extensive experiments on 16 datasets demonstrate that DOC improves adversarial robustness under various attacks while maintaining competitive clean accuracy.

PDF Details DOI

EAAI Journal 2025 Journal Article

Adaptive gradient-aware neural dynamics: Towards fast and accurate solutions for dynamic convex optimization

Chengze Jiang
Aiping Ye
Huiting He
Xiuchun Xiao
Cong Lin

Constrained Dynamic Convex Optimization (CDCO) represents a core challenge in many engineering applications, where the objective is to minimize a time-varying cost function subject to dynamically evolving constraints. While recent neural network-based methods have demonstrated potential in addressing CDCO, they still suffer from limitations in convergence and solution accuracy, which restrict their effectiveness in real-world deployments. To overcome these challenges, we propose an Adaptive Gradient-Aware Neural Dynamics (AGAND). As an artificial intelligence model derived from a branch of Hopfield networks, AGAND integrates a gradient-aware term with a time derivative term to enhance solution performance. Our AGAND uses gradient information and time-derivative data to achieve faster and more accurate solutions for CDCO. Besides, the convergence of the model is further boosted by introducing state-aware coefficient with gradient feedback mechanism. Theoretical analysis demonstrates the global convergence of the AGAND, along with a detailed complexity assessment. To further adapt AGAND for practical deployment, a discretization scheme is proposed to facilitate implementation on digital hardware. Comparative experiments with state-of-the-art methods illustrate the competitiveness of our AGAND in terms of convergence and accuracy, achieving Average Steady-state Residual Error (ASSRE) of 3. 10 × 1 0 − 3 and Convergence Time (CT) of 0. 04 s. Finally, a robot kinematics scheme and hyperspectral image target detection are formulated on the basis of our AGAND, demonstrating the feasibility and practical utility of the AGAND in real-world engineering problems.

Details DOI

AAAI Conference 2024 Conference Paper

Taxonomy Driven Fast Adversarial Training

Kun Tong
Chengze Jiang
Jie Gui
Yuan Cao

Adversarial training (AT) is an effective defense method against gradient-based attacks to enhance the robustness of neural networks. Among them, single-step AT has emerged as a hotspot topic due to its simplicity and efficiency, requiring only one gradient propagation in generating adversarial examples. Nonetheless, the problem of catastrophic overfitting (CO) that causes training collapse remains poorly understood, and there exists a gap between the robust accuracy achieved through single- and multi-step AT. In this paper, we present a surprising finding that the taxonomy of adversarial examples reveals the truth of CO. Based on this conclusion, we propose taxonomy driven fast adversarial training (TDAT) which jointly optimizes learning objective, loss function, and initialization method, thereby can be regarded as a new paradigm of single-step AT. Compared with other fast AT methods, TDAT can boost the robustness of neural networks, alleviate the influence of misclassified examples, and prevent CO during the training process while requiring almost no additional computational and memory resources. Our method achieves robust accuracy improvement of 1.59%, 1.62%, 0.71%, and 1.26% on CIFAR-10, CIFAR-100, Tiny ImageNet, and ImageNet-100 datasets, when against projected gradient descent PGD10 attack with perturbation budget 8/255. Furthermore, our proposed method also achieves state-of-the-art robust accuracy against other attacks. Code is available at https://github.com/bookman233/TDAT.

PDF Details DOI