Author name cluster

Zhijie Wang

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.

9 papers

2 author rows

AAAI Conference 2026 Conference Paper

Beyond Single-Point Perturbation: A Hierarchical, Manifold-Aware Approach to Diffusion Attacks

Zhijie Wang
Lin Wang
Zhenyu Wen
Cong Wang

Latent Diffusion Models have become a powerful tool for generating high-fidelity unrestricted adversarial examples. However, the existing methods typically perturb only the initial latent or rely on prompt engineering, which is ill-suited to the iterative nature of the diffusion process, plus optimization instability due to external text prompts and cumulative drift that push the adversarial images off the data manifold. In this paper, we propose a hierarchical attack framework that operates in alignment with the model's generative manifold and leverages intermediate denoising states to maximize attack transferability and visual fidelity. Extensive experiments show that the proposed attack improves adversarial transferability by 10-20% against a diverse set of normally-trained models and achieves over 10.5% higher success rate against adversarially-defended models, while simultaneously enhancing visual quality by 1.0-1.2 FID reduction and 16.7% LPIPS improvements.

PDF Details DOI

EAAI Journal 2025 Journal Article

Confidence-driven under-sampling decision forest for imbalanced credit scoring

Meng Xia
Zhijie Wang
Xingyu Lan
Wanan Liu
Jiawei Wu

Credit scoring is vital for assessing financial trustworthiness. Ensemble frameworks and resampling-based strategies have great potential in handling imbalanced credit scoring. However, resampling-based ensemble learning can potentially distort the original credit dataset distribution, thereby posing challenges in optimizing each base predictor. Additionally, achieving balance in the entire sample space with one-step resampling is challenging and can lead to information loss. To tackle these issues, a confidence-driven under-sampling decision forest (CUDF) is proposed. CUDF is established as an ensemble approach, harnessing the collective power of multiple decision trees; in addition, a confidence-driven under-sampling (CDUS) mechanism is incorporated into CUDF, enabling each base learner within CUDF to become an imbalance-aware predictor; besides, a weighted aggregation scheme is designed for CUDF enable precise fine-tuning of weight allocation for each base classifier, facilitating tailored adjustments to maximize their contributions for imbalanced credit scoring. Empirical findings substantiate that the CUDF surpasses the performance of both traditional credit scoring models and imbalanced credit scoring models across four distinct credit scoring datasets. The algorithm not only amplifies the precision in categorizing risky borrower classes but also demonstrates notable efficiency in managing extensive imbalanced credit scoring datasets. Consequently, CUDF exhibits promising potential for applications in addressing imbalanced data challenges within the domain of credit scoring.

Details DOI

ICLR Conference 2025 Conference Paper

RaSA: Rank-Sharing Low-Rank Adaptation

Zhiwei He 0002
Zhaopeng Tu
Xing Wang 0007
Xingyu Chen
Zhijie Wang
Jiahao Xu
Tian Liang
Wenxiang Jiao

Low-rank adaptation (LoRA) has been prominently employed for parameter-efficient fine-tuning of large language models (LLMs). However, the limited expressive capacity of LoRA, stemming from the low-rank constraint, has been recognized as a bottleneck, particularly in rigorous tasks like code generation and mathematical reasoning. To address this limitation, we introduce Rank-Sharing Low-Rank Adaptation (RaSA), an innovative extension that enhances the expressive capacity of LoRA by leveraging partial rank sharing across layers. By forming a shared rank pool and applying layer-specific weighting, RaSA effectively increases the number of ranks without augmenting parameter overhead. Our theoretically grounded and empirically validated approach demonstrates that RaSA not only maintains the core advantages of LoRA but also significantly boosts performance in challenging code and math tasks. Code, data and scripts are available at: https://github.com/zwhe99/RaSA.

Details

ICML Conference 2024 Conference Paper

In-context Learning on Function Classes Unveiled for Transformers

Zhijie Wang
Bo Jiang
Shuai Li 0010

Transformer-based neural sequence models exhibit a remarkable ability to perform in-context learning. Given some training examples, a pre-trained model can make accurate predictions on an unseen input. This paper studies why transformers can learn different types of function classes in-context. We first show by construction that there exists a family of transformers (with different activation functions) that implement approximate gradient descent on the parameters of neural networks, and we provide an upper bound for the number of heads, hidden dimensions, and layers of the transformer. We also show that a transformer can learn linear functions, the indicator function of a unit ball, and smooth functions in-context by learning neural networks that approximate them. The above instances mainly focus on a transformer pre-trained on single tasks. We also prove that when pre-trained on two tasks: linear regression and classification, a transformer can make accurate predictions on both tasks simultaneously. Our results move beyond linearity in terms of in-context learning instances and provide a comprehensive understanding of why transformers can learn many types of function classes through the bridge of neural networks.

Details

NeurIPS Conference 2024 Conference Paper

Leveraging Drift to Improve Sample Complexity of Variance Exploding Diffusion Models

Ruofeng Yang
Zhijie Wang
Bo Jiang
Shuai Li

Variance exploding (VE) based diffusion models, an important class of diffusion models, have shown state-of-the-art (SOTA) performance. However, only a few theoretical works analyze VE-based models, and those works suffer from a worse forward convergence rate $1/\text{poly}(T)$ than the $\exp{(-T)}$ of variance preserving (VP) based models, where $T$ is the forward diffusion time and the rate measures the distance between forward marginal distribution $q_T$ and pure Gaussian noise. The slow rate is due to the Brownian Motion without a drift term. In this work, we design a new drifted VESDE forward process, which allows a faster $\exp{(-T)}$ forward convergence rate. With this process, we achieve the first efficient polynomial sample complexity for a series of VE-based models with reverse SDE under the manifold hypothesis. Furthermore, unlike previous works, we allow the diffusion coefficient to be unbounded instead of a constant, which is closer to the SOTA models. Besides the reverse SDE, the other common reverse process is the probability flow ODE (PFODE) process, which is deterministic and enjoys faster sample speed. To deepen the understanding of VE-based models, we consider a more general setting considering reverse SDE and PFODE simultaneously, propose a unified tangent-based analysis framework, and prove the first quantitative convergence guarantee for SOTA VE-based models with reverse PFODE. We also show that the drifted VESDE can balance different error terms and improve generated samples without training through synthetic and real-world experiments.

PDF Details DOI

IJCAI Conference 2023 Conference Paper

Accurate MRI Reconstruction via Multi-Domain Recurrent Networks

Jinbao Wei
Zhijie Wang
Kongqiao Wang
Li Guo
Xueyang Fu
Ji Liu
Xun Chen

In recent years, deep convolutional neural networks (CNNs) have become dominant in MRI reconstruction from undersampled k-space. However, most existing CNNs methods reconstruct the undersampled images either in the spatial domain or in the frequency domain, and neglecting the correlation between these two domains. This hinders the further reconstruction performance improvement. To tackle this issue, in this work, we propose a new multi-domain recurrent network (MDR-Net) with multi-domain learning (MDL) blocks as its basic units to reconstruct the undersampled MR image progressively. Specifically, the MDL block interactively processes the local spatial features and the global frequency information to facilitate complementary learning, leading to fine-grained features generation. Furthermore, we introduce an effective frequency-based loss to narrow the frequency spectrum gap, compensating for over-smoothness caused by the widely used spatial reconstruction loss. Extensive experiments on public fastMRI datasets demonstrate that our MDR-Net consistently outperforms other competitive methods and is able to provide more details.

PDF Details DOI

YNIMG Journal 2020 Journal Article

Cortical hemodynamic responses induced by low-intensity transcranial ultrasound stimulation of mouse cortex

Yi Yuan
Zhijie Wang
Mengyang Liu
Shy Shoham

Ultrasound-mediated neuromodulation is emerging as a key technology for targeted noninvasive brain stimulation, but key insights into its effects and dose-response characteristics are still missing. The purpose of this study is to systematically evaluate the effect of low-intensity transcranial ultrasound stimulation (TUS) on complementary aspects of cerebral hemodynamic. We simultaneously record the EMG signal, local field potential (LFP) and cortical blood flow (CBF) using electrophysiological recording and laser speckle contrast imaging under ultrasound stimulation to simultaneously monitor motor responses, neural activities and hemodynamic changes during the application of low-intensity TUS in mouse motor cortex, using excitation pulses which caused whisker and tail movement. Our experimental results demonstrate interdependent TUS-induced motor, neural activity and hemodynamic responses that peak approximately 0.55s, 1.05s and 2.5s after TUS onset, respectively, and show a linear coupling relationship between their respective varying response amplitudes to repeated stimuli. We also found monotonic dose-response parametric relations of the CBF peak value increase as a function of stimulation intensity and duration, while stimulus duty-cycle had only a weak effect on peak responses. These findings demonstrate that TUS induces a change in cortical hemodynamics and LSCI provide a high temporal resolution view of these changes.

Details DOI

IJCAI Conference 2018 Conference Paper

Image-level to Pixel-wise Labeling: From Theory to Practice

Tiezhu Sun
Wei Zhang
Zhijie Wang
Lin Ma
Zequn Jie

Conventional convolutional neural networks (CNNs) have achieved great success in image semantic segmentation. Existing methods mainly focus on learning pixel-wise labels from an image directly. In this paper, we advocate tackling the pixel-wise segmentation problem by considering the image-level classification labels. Theoretically, we analyze and discuss the effects of image-level labels on pixel-wise segmentation from the perspective of information theory. In practice, an end-to-end segmentation model is built by fusing the image-level and pixel-wise labeling networks. A generative network is included to reconstruct the input image and further boost the segmentation model training with an auxiliary loss. Extensive experimental results on benchmark dataset demonstrate the effectiveness of the proposed method, where good image-level labels can significantly improve the pixel-wise segmentation accuracy.

PDF Details

IROS Conference 2006 Conference Paper

Humanoid Motion Design Considering Rhythm Based on Human Motion Capture

Lige Zhang
Qiang Huang
Shusheng Lv
You Shi
Zhijie Wang
Ali Raza Jafri

This paper explores the method of designing humanoid motion considering rhythm based on human motion capture. Captured human data must be adapted for the humanoid because its kinematics and dynamics differ from those of the human actor. On the other hand, it is expected that humanoid movements are highly similar to those of the human actor. In this paper, first the motion of the human actor is segmented into primitive motions. Then, the kinematics constraints and stability adjustment method are formulated. Next, the similarity evaluation considering rhythm is discussed, and the method to derive humanoid motion with high similarity, and satisfying kinematic constraints and dynamic stability, is presented. Finally, the effectiveness of the proposed method is illustrated by the experiment of Chinese Kungfu "sword" motion using our developed humanoid robot BHR-2 with 32 DOF

Details