Search

Your search keyword '"Wang, Yuji"' showing total 717 results
Start Over Author "Wang, Yuji"
717 results on '"Wang, Yuji"'

1. FrameBridge: Improving Image-to-Video Generation with Bridge Models

Author

Wang, Yuji, Chen, Zehua, Chen, Xiaoyu, Zhu, Jun, and Chen, Jianfei

Subjects
Computer Science - Computer Vision and Pattern Recognition, Computer Science - Artificial Intelligence, Computer Science - Machine Learning
Abstract

Image-to-video (I2V) generation is gaining increasing attention with its wide application in video synthesis. Recently, diffusion-based I2V models have achieved remarkable progress given their novel design on network architecture, cascaded framework, and motion representation. However, restricted by their noise-to-data generation process, diffusion-based methods inevitably suffer the difficulty to generate video samples with both appearance consistency and temporal coherence from an uninformative Gaussian noise, which may limit their synthesis quality. In this work, we present FrameBridge, taking the given static image as the prior of video target and establishing a tractable bridge model between them. By formulating I2V synthesis as a frames-to-frames generation task and modelling it with a data-to-data process, we fully exploit the information in input image and facilitate the generative model to learn the image animation process. In two popular settings of training I2V models, namely fine-tuning a pre-trained text-to-video (T2V) model or training from scratch, we further propose two techniques, SNR-Aligned Fine-tuning (SAF) and neural prior, which improve the fine-tuning efficiency of diffusion-based T2V models to FrameBridge and the synthesis quality of bridge-based I2V models respectively. Experiments conducted on WebVid-2M and UCF-101 demonstrate that: (1) our FrameBridge achieves superior I2V quality in comparison with the diffusion counterpart (zero-shot FVD 83 vs. 176 on MSR-VTT and non-zero-shot FVD 122 vs. 171 on UCF-101); (2) our proposed SAF and neural prior effectively enhance the ability of bridge-based I2V models in the scenarios of fine-tuning and training from scratch. Demo samples can be visited at: https://framebridge-demo.github.io/.

Published
2024

2. Robust Multimodal Learning via Representation Decoupling

Author

Wei, Shicai, Luo, Yang, Wang, Yuji, and Luo, Chunbo

Subjects
Computer Science - Computer Vision and Pattern Recognition, Computer Science - Artificial Intelligence
Abstract

Multimodal learning robust to missing modality has attracted increasing attention due to its practicality. Existing methods tend to address it by learning a common subspace representation for different modality combinations. However, we reveal that they are sub-optimal due to their implicit constraint on intra-class representation. Specifically, the sample with different modalities within the same class will be forced to learn representations in the same direction. This hinders the model from capturing modality-specific information, resulting in insufficient learning. To this end, we propose a novel Decoupled Multimodal Representation Network (DMRNet) to assist robust multimodal learning. Specifically, DMRNet models the input from different modality combinations as a probabilistic distribution instead of a fixed point in the latent space, and samples embeddings from the distribution for the prediction module to calculate the task loss. As a result, the direction constraint from the loss minimization is blocked by the sampled representation. This relaxes the constraint on the inference representation and enables the model to capture the specific information for different modality combinations. Furthermore, we introduce a hard combination regularizer to prevent DMRNet from unbalanced training by guiding it to pay more attention to hard modality combinations. Finally, extensive experiments on multimodal classification and segmentation tasks demonstrate that the proposed DMRNet outperforms the state-of-the-art significantly., Comment: ECCV2024 17 pages

Published
2024

3. Fight Back Against Jailbreaking via Prompt Adversarial Tuning

Author

Mo, Yichuan, Wang, Yuji, Wei, Zeming, and Wang, Yisen

Subjects
Computer Science - Machine Learning, Computer Science - Artificial Intelligence, Computer Science - Computation and Language, Computer Science - Cryptography and Security
Abstract

While Large Language Models (LLMs) have achieved tremendous success in various applications, they are also susceptible to jailbreaking attacks. Several primary defense strategies have been proposed to protect LLMs from producing harmful information, mostly focusing on model fine-tuning or heuristical defense designs. However, how to achieve intrinsic robustness through prompt optimization remains an open problem. In this paper, motivated by adversarial training paradigms for achieving reliable robustness, we propose an approach named Prompt Adversarial Tuning (PAT) that trains a prompt control attached to the user prompt as a guard prefix. To achieve our defense goal whilst maintaining natural performance, we optimize the control prompt with both adversarial and benign prompts. Comprehensive experiments show that our method is effective against both grey-box and black-box attacks, reducing the success rate of advanced attacks to nearly 0%, while maintaining the model's utility on the benign task and incurring only negligible computational overhead, charting a new perspective for future explorations in LLM security. Our code is available at https://github.com/PKU-ML/PAT.

Published
2024

4. Robust Multimodal Learning via Representation Decoupling

Author

Wei, Shicai, Luo, Yang, Wang, Yuji, Luo, Chunbo, Goos, Gerhard, Series Editor, Hartmanis, Juris, Founding Editor, Bertino, Elisa, Editorial Board Member, Gao, Wen, Editorial Board Member, Steffen, Bernhard, Editorial Board Member, Yung, Moti, Editorial Board Member, Leonardis, Aleš, editor, Ricci, Elisa, editor, Roth, Stefan, editor, Russakovsky, Olga, editor, Sattler, Torsten, editor, and Varol, Gül, editor

Published
2025
Full Text
View/download PDF

6. Phylodynamics unveils invading and diffusing patterns of dengue virus serotype-1 in Guangdong, China from 1990 to 2019 under a global genotyping framework

Author

Zhao, Lingzhai, Guo, Xiang, Li, Liqiang, Jing, Qinlong, Ma, Jinmin, Xie, Tian, Lin, Dechun, Li, Li, Yin, Qingqing, Wang, Yuji, Zhang, Xiaoqing, Li, Ziyao, Liu, Xiaohua, Hu, Tian, Hu, Minling, Ren, Wenwen, Li, Jun, Peng, Jie, Yu, Lei, Peng, Zhiqiang, Hong, Wenxin, Leng, Xingyu, Luo, Lei, Ngobeh, Jone Jama Kpanda, Tang, Xiaoping, Wu, Rangke, Zhao, Wei, Shi, Benyun, Liu, Jiming, Yang, Zhicong, Chen, Xiao-Guang, Zhou, Xiaohong, and Zhang, Fuchun

Published
2024
Full Text
View/download PDF

14. Therapeutic potential of arginine deprivation therapy for gliomas: a systematic review of the existing literature

Author

Chen Yuxiao, Wang Jiachen, Lan Yanjie, Li Shenglan, Wang Yuji, and Li Wenbin

Subjects
glioma, arginine deprivation therapy, tumor metabolism, combination therapy, ADI-PEG20, BET-100, Therapeutics. Pharmacology, RM1-950
Abstract

BackgroundArginine deprivation therapy (ADT) hinders glioma cells’ access to nutrients by reducing peripheral blood arginine, showing great efficacy in various studies, which suggests it as a potentially promising treatment for glioma. The aim of this systematic review was to explore the mechanism of ADT for gliomas, the therapeutic effect based on existing research, and possible combination therapies.MethodsWe performed a systematic literature review of PubMed, ScienceDirect and Web of Science databases according to PRISMA guidelines, searching for articles on the efficacy of ADT in glioma.ResultsWe identified 17 studies among 786 search results, among which ADT therapy mainly based on Arginine free condition, Arginine Deiminase and Arginase, including three completed clinical trials. ADT therapy has shown promising results in vivo and in vitro, with its safety confirmed in clinical trials. In the early phase of treatment, glioblastoma (GBM) cells develop protective mechanisms of stress and autophagy, which eventually evolve into caspase dependent apoptosis or senescence, respectively. The immunosuppressive microenvironment is also altered by arginine depletion, such as the transformation of microglia into a pro-inflammatory phenotype and the activation of T-cells. Thus, ADT therapy demonstrates glioma-killing effect in the presence of a combination of mechanisms. In combination with various conventional therapies and investigational drugs such as radiotherapy, temozolomide (TMZ), cyclin-dependent kinase inhibitors (CDK) inhibitors and autophagy inducers, ADT therapy has been shown to be more effective. However, the phenomenon of drug resistance due to re-expression of ASS1 rather than stem cell remains to be investigated.ConclusionDespite the paucity of studies in the literature, the available data demonstrate the therapeutic potential of arginine deprivation therapy for glioma and encourage further research, especially the exploration of its combination therapies and the extrapolation of what we know about the effects and mechanisms of ADT from other tumors to glioma.

Published
2024
Full Text
View/download PDF

22. Field investigation combined with modeling uncovers the ecological heterogeneity of Aedes albopictus habitats for strategically improving systematic management during urbanization

Author

Guo, Xiang, Luo, Lei, Long, Yuxiang, Teng, Pingying, Wei, Yuehong, Xie, Tian, Li, Li, Yin, Qingqing, Li, Ziyao, Wang, Yuji, He, Jiejun, Ji, Xiatian, Zhou, Huasheng, Zhang, Xiaofan, Chen, Shigang, Zhou, Yezhen, Xu, Kaihui, Liang, Guancong, Kuang, Haocheng, Gao, Yuting, Liu, Xiaohua, Luo, Luting, Ding, Lin, Li, Yiji, Liu, Zhuanzhuan, Zhou, Tengfei, Lai, Zetian, Su, Xinghua, Guo, Yuyan, Li, Chenying, Xie, Lihua, Li, Minqing, Wu, Xinglong, Huang, Jianhao, Su, Weicong, Pan, Yicheng, Hu, Wei, Zhou, Dongrui, Li, Chunv, Gui, Juan, Ma, Jiazhi, Feng, Xiaoli, Zhu, Minyi, Zhong, Shangbin, Chen, Fan, Zeng, Huanchao, Wu, Yingxian, Wang, Chen, Li, Shukai, Wang, Qi, Wang, Xueyi, Zhou, Yekai, Ling, Jianxun, Liu, Yingjie, Wu, Shang, Li, Zhiwei, Zhong, Minghui, Wei, Wenxia, Xie, Lixian, Xu, Xianli, Huang, Hehai, Yang, Guilan, Liu, Yan, Liang, Siting, Wu, Yingxia, Zhang, Deyu, Xu, Changqing, Wang, Jie, Wang, Chunmei, Wu, Rangke, Yang, Zhicong, Chen, Xiao-Guang, and Zhou, Xiaohong

Published
2023
Full Text
View/download PDF

45. Studious Bob Fight Back Against Jailbreaking via Prompt Adversarial Tuning

Author

Mo, Yichuan, Wang, Yuji, Wei, Zeming, Wang, Yisen, Mo, Yichuan, Wang, Yuji, Wei, Zeming, and Wang, Yisen

Abstract

Although Large Language Models (LLMs) have achieved tremendous success in various applications, they are also susceptible to certain prompts that can induce them to bypass built-in safety measures and provide dangerous or illegal content, a phenomenon known as jailbreak. To protect LLMs from producing harmful information, various defense strategies are proposed, with most focusing on content filtering or adversarial training of models. In this paper, we propose an approach named Prompt Adversarial Tuning (PAT) to train a defense control mechanism, which is then embedded as a prefix to user prompts to implement our defense strategy. We design a training process similar to adversarial training to achieve our optimized goal, alternating between updating attack and defense controls. To our knowledge, we are the first to implement defense from the perspective of prompt tuning. Once employed, our method will hardly impact the operational efficiency of LLMs. Experiments show that our method is effective in both black-box and white-box settings, reducing the success rate of advanced attacks to nearly 0 while maintaining the benign answer rate of 80% to simple benign questions. Our work might potentially chart a new perspective for future explorations in LLM security.

Published
2024

48. RGD-modified ZIF-8 nanoparticles as a drug carrier for MR imaging and targeted drug delivery in myocardial infarction.

Author

Li, Yingxu, Tuerhan, Maisituremu, Li, Bing, Chen, Shuangling, Wang, Yuji, and Zheng, Yuanyuan

Abstract

Aim: A multifunctional nanoplatform has been developed to enhance the targeting capability and biosafety of drug/siRNA for better diagnosis and treatment of myocardial infarction (MI). Materials & methods: The nanoplatform's chemical properties, biodistribution, cardiac magnetic resonance imaging (MRI) capabilities, therapeutic effects and biocompatibility were investigated. Results: The nanoplatform exhibited MI-targeting properties and pH-sensitivity, allowing for effective cardiac MRI and delivery of drugs to the infarcted myocardium. The GCD/Qt@ZIF-RGD demonstrated potential as a reliable MRI probe for MI diagnosis. Moreover, the GCD/si-SHP1/Qt@ZIF-RGD effectively suppressed SHP-1 expression, increased pro-angiogenesis gene expression and reduced cell apoptosis in HUVECs exposed to hypoxia/reoxygenation. Conclusion: Our newly developed multifunctional drug delivery system shows promise as a nanoplatform for both the diagnosis and treatment of MI. Graphical Abstract Article highlights The BSA-RGD-modified ZIF-8 nanoplatform is constructed. The nanoplatform encapsulates quercetin and SHP-1 siRNA as the therapeutic agents, and GCD as the MRI contrast agent. The nanoplatform is infarct-targeting and pH-sensitive, resulting in efficient drug/siRNA delivery. The GCD/si-SHP1/Qt@ZIF-RGD nanoparticles produce efficient gene silencing of SHP-1 in vitro and in vivo. The GCD/si-Con/Qt@ZIF-RGD promote angiogenesis and alleviate cell apoptosis on HUVECs under H/R. The GCD/si-SHP1/Qt@ZIF-RGD display better performance on promoting angiogenesis and alleviating cell apoptosis on HUVECs under H/R. The GCD/Qt@ZIF-RGD can serve as a promising MRI probe for MI imaging. The GCD/si-SHP1/Qt@ZIF-RGD nanoparticles show good biocompatibility. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

49. PAD4 Inhibitor‐Functionalized Layered Double Hydroxide Nanosheets for Synergistic Sonodynamic Therapy/Immunotherapy Of Tumor Metastasis.

Author

Zhu, Di, Lu, Yu, Yang, Shuqing, Hu, Tingting, Tan, Chaoliang, Liang, Ruizheng, and Wang, Yuji

Subjects
LAYERED double hydroxides, METASTASIS, HYDROXIDES, ARGININE deiminase, CANCER invasiveness, IMMUNOTHERAPY, ROOT-tubercles
Abstract

Sonodynamic therapy (SDT) is demonstrated to trigger the systemic immune response of the organism and facilitate the treatment of metastatic tumors. However, SDT‐mediated neutrophil extracellular traps (NETs) formation can promote tumor cell spread, thus weakening the therapeutic effectiveness of metastatic tumors. Herein, the amorphous CoW‐layered double hydroxide (a‐CoW‐LDH) nanosheets are functionalized with a peptidyl arginine deiminase 4 (PAD4) inhibitor, i.e., YW3‐56, to construct a multifunctional nanoagent (a‐LDH@356) for synergistic SDT/immunotherapy. Specifically, a‐CoW‐LDH nanosheets can act as a sonosensitizer to generate abundant reactive oxygen species (ROS) under US irradiation. After loading with YW3‐56, a‐LDH@356 plus US irradiation not only effectively induces ROS generation and immunogenic cell death, but also inhibits the elevation of citrullinated histone H3 (H3cit) and the release of NETs, enabling a synergistic enhancement of anti‐tumor metastasis effect. Using 4T1 tumor model, it is demonstrated that combining a‐CoW‐LDH with YW3‐56 stimulates an anti‐tumor response by upregulating the proportion of immune‐activated cells and inducing polarization of M1 macrophages, and inhibits immune escape by downregulating the expression of PD‐1 on immune cells under US irradiation, which not only arrests primary tumor progression with a tumor inhibition rate of 69.5% but also prevents tumor metastasis with the least number of lung metastatic nodules. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

50. The Effect of Basic Ligands and Alkenes on the Regioselectivity of C−H Additions of Tertiary Amines to Alkenes.

Author

Li, Da, Wang, Yuji, Wang, Shiyu, and Dong, Shunxi

Subjects
*TERTIARY amines, *ALKENES, *AMINE derivatives, *ALKYLATION, *DEUTERIUM, *ANILINE, *SCANDIUM
Abstract

Highly regioselective C−H alkylation reactions of tertiary anilines and tertiary alkyl amines with simple alkenes have been achieved by the use of imidazolin‐2‐iminato scandium alkyl complexes. This protocol provided an efficient and atom‐economical route to structurally diverse tertiary amine derivatives. The basic ligand, a coordinating THF in the catalyst and the substitution of alkene substrates were found to switch the regioselectivity of the C−H alkylation reactions of tertiary anilines presumably due to the generation of different types of catalytically active species or the formation of relatively stable intermediates. On the basis of the deuterium labeling experiments and KIE experiments, possible catalytical cycles were provided to understand the reaction mechanism as well as the regioselectivity. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results