Zhipeng Wei

Changchun University of Science and Technology

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Trends

Author Order

Document Type

Co-Authors

Xuan Fang

Changchun University of Science and Technology

114

Jilong Tang

Changchun University of Science and Technology

Xiaohua Wang

Changchun University of Science and Technology

Dan Fang

Changchun University of Science and Technology

Dengkui Wang

Changchun University of Science and Technology

Xueying Chu

Changchun University of Science and Technology

Rui Chen

Southern University of Science and Technology

Jinhua Li

Hubei University

Fengyuan Lin

Huazhong University of Science and Technology

Xiaohua Wang

Xi'an Jiaotong University

Cooperative Institutions

Chinese Academy of Sciences

316

Changchun University of Science and Technology

213

University of Chinese Academy of Sciences

161

China Medical University

104

China Medical University Hospital

Jilin University

University of Science and Technology of China

National Yang Ming Chiao Tung University

Zhejiang University

Changchun Institute of Optics, Fine Mechanics and Physics

Author Statistics

Papers

Citation

H-Index

i-10 index

Research Field

Reliable and Efficient Concept Erasure of Text-to-Image Diffusion Models

arXiv (Cornell University) (2024)

Chao Gong Kai Chen Zhipeng Wei Jingjing Chen Yu–Gang Jiang

Text-to-image models encounter safety issues, including concerns related to copyright and Not-Safe-For-Work (NSFW) content. Despite several methods have been proposed for erasing inappropriate concepts from diffusion models, they often exhibit incomplete erasure, consume a lot of computing resources, and inadvertently damage generation ability. In this work, we introduce Reliable and Efficient Concept Erasure (RECE), a novel approach that modifies the model in 3 seconds without necessitating additional fine-tuning. Specifically, RECE efficiently leverages a closed-form solution to derive new target embeddings, which are capable of regenerating erased concepts within the unlearned model. To mitigate inappropriate content potentially represented by derived embeddings, RECE further aligns them with harmless concepts in cross-attention layers. The derivation and erasure of new representation embeddings are conducted iteratively to achieve a thorough erasure of inappropriate concepts. Besides, to preserve the model's generation ability, RECE introduces an additional regularization term during the derivation process, resulting in minimizing the impact on unrelated concepts during the erasure process. All the processes above are in closed-form, guaranteeing extremely efficient erasure in only 3 seconds. Benchmarking against previous approaches, our method achieves more efficient and thorough erasure with minor damage to original generation ability and demonstrates enhanced robustness against red-teaming tools. Code is available at \url{https://github.com/CharlesGong12/RECE}.

Erasure

10.48550/arxiv.2407.12383

Cite

Citations (0)

[Effects of annealing temperature on the structure and optical properties of ZnMgO films prepared by atom layer deposition].

PubMed (2014)

Dongxiao Sun Jinhua Li Xuan Fang Xinying Chen Fang Fang

In the present paper, we report the research on the effects of annealing temperature on the crystal quality and optical properties of ZnMgO films deposited by atom layer deposition(ALD). ZnMgO films were prepared on quartz substrates by ALD and then some of the samples were treated in air ambient at different annealing temperature. The effects of annealing temperature on the crystal quality and optical properties of ZnMgO films were characterized by X-ray diffraction (XRD), photoluminescence (PL) and ultraviolet-visible (UV-Vis) absorption spectra. The XRD results showed that the crystal quality of ZnMgO films was significantly improved when the annealing temperature was 600 degrees C, meanwhile the intensity of(100) diffraction peak was the strongest. Combination of PL and UV-Vis absorption measurements showed that it can strongly promote the Mg content increasing in ZnMgO films and increase the band gap of films. So the results illustrate that suitable annealing temperature can effectively improve the crystal quality and optical properties of ZnMgO films.

Crystal (programming language)

Source

Cite

Citations (0)

Attacking Video Recognition Models with Bullet-Screen Comments

arXiv (Cornell University) (2021)

Kai Chen Zhipeng Wei Jingjing Chen Zuxuan Wu Yu–Gang Jiang

Recent research has demonstrated that Deep Neural Networks (DNNs) are vulnerable to adversarial patches which introduce perceptible but localized changes to the input. Nevertheless, existing approaches have focused on generating adversarial patches on images, their counterparts in videos have been less explored. Compared with images, attacking videos is much more challenging as it needs to consider not only spatial cues but also temporal cues. To close this gap, we introduce a novel adversarial attack in this paper, the bullet-screen comment (BSC) attack, which attacks video recognition models with BSCs. Specifically, adversarial BSCs are generated with a Reinforcement Learning (RL) framework, where the environment is set as the target model and the agent plays the role of selecting the position and transparency of each BSC. By continuously querying the target models and receiving feedback, the agent gradually adjusts its selection strategies in order to achieve a high fooling rate with non-overlapping BSCs. As BSCs can be regarded as a kind of meaningful patch, adding it to a clean video will not affect people' s understanding of the video content, nor will arouse people' s suspicion. We conduct extensive experiments to verify the effectiveness of the proposed method. On both UCF-101 and HMDB-51 datasets, our BSC attack method can achieve about 90\% fooling rate when attacking three mainstream video recognition models, while only occluding \textless 8\% areas in the video. Our code is available at https://github.com/kay-ck/BSC-attack.

Code (set theory)

Deep Neural Networks

Mainstream

10.48550/arxiv.2110.15629

Cite

Citations (4)

Facile synthesis via a solvent molecular template and formation mechanism of uniform zinc antimony nanorods

Journal of Materials Science Materials in Electronics (2018)

Haoran Zhang Xinwei Wang Xinyang Cai Dengkui Wang Jilong Tang

Nanorod

10.1007/s10854-018-9102-4

Cite

Citations (1)

PEALD堆積ALNの構造と表面特性に及ぼす温度の影響を【JST・京大機械翻訳】

Hongwai yu Jiguang Gongcheng (2016)

Chen Fang Xuan Fang Shuangpeng Wang Shouzhu Niu Fang Fang

Source

Cite

Citations (0)

Electrochemical Biosensor with pH Regulation of CNTs/HRP Multilayer for Phenols

Analytical Sciences (2011)

Shaoming Yang Aihua Huang Dan Jiang Zhipeng Wei Longzhen Zheng

Amperometry

Catechol

Horseradish peroxidase

10.2116/analsci.27.415

Cite

Citations (3)

Surface passivation of GaAs using atomic layer deposition grown MgO

Materials Research Express (2015)

Xian Gao Dan Fang Xuan Fang Jilong Tang Fang Fang

In this paper, different MgO films were deposited on a GaAs substrate by atom layer deposition (ALD), which can passivate the surface state of GaAs. From XPS results, the band structure of MgO/GaAs can be indexed to type I. Based on band alignment, the excitons could not diffuse and created nonradiative centers, which were all limited and then recombined at the MgO/GaAs interface, thus the emission intensity of the 100 nm MgO film coated sample was about 3 times higher than the uncoated sample.

Passivation

Deposition

10.1088/2053-1591/2/9/095902

Cite

Citations (7)

Zn or O? An Atomic Level Comparison on Antibacterial Activities of Zinc Oxides

Chemistry - A European Journal (2016)

Fen Yu Xuan Fang Huimin Jia Miaoxing Liu Xiaotong Shi

Abstract For the first time, the influence of different types of atoms (Zn and O) on the antibacterial activities of nanosized ZnO was quantitatively evaluated with the aid of a 3D‐printing‐manufactured evaluation system. Two different outermost atomic layers were manufactured separately by using an ALD (atomic layer deposition) method. Interestingly, we found that each outermost atomic layer exhibited certain differences against gram‐positive or gram‐negative bacterial species. Zinc atoms as outermost layer (ZnO−Zn) showed a more pronounced antibacterial effect towards gram‐negative E. coli ( Escherichia coli ), whereas oxygen atoms (ZnO−O) showed a stronger antibacterial activity against gram‐positive S. aureus ( Staphylococcus aureus ). A possible antibacterial mechanism has been comprehensively discussed from different perspectives, including Zn 2+ concentrations, oxygen vacancies, photocatalytic activities and the DNA structural characteristics of different bacterial species.

Gram

10.1002/chem.201601018

Cite

Citations (39)

Ethylene for carbon doping of GaN by atmospheric pressure metal organic chemical vapor deposition

Materials Letters (2023)

Li Zhang Zhongyuan Dong Xuguang Deng Xin Zhou Kun Xu

Trimethylgallium

Carbon fibers

10.1016/j.matlet.2023.134475

Cite

Citations (0)