Mail: zzy1210@pku.edu.cn
Zhiyuan Zhang (张之远) received the Bachelor’s degree of Science in Computer Science from Peking University in 2019 and received the Doctoral degree of Science in Computer Science from Peking University in 2024, supervised by Prof. Xu Sun (孙栩). His research interests include deep learning, natural language processing and the application to quantitative finance.
Sept. 2019 - Jul. 2024
PhD student, majoring in Computer Software and Theory, at School of Computer Science, Peking University.
Advisor: Xu Sun (孙栩). GPA: 3.85/4.00, Rank: 1.
Sept. 2015 - Jul. 2019
Bachelor, majoring in Science in Computer Science, at School of EECS, Peking University.
GPA: 3.76/4.00, Rank: 6/204.
Zhiyuan Zhang has 30 papers, including 16 first-author papers. The papers have been cited 1000+ times in Google Scholar up to 2024. Followings are selected publications (#: Equal Contribution).
Links: [Google Scholar], [OpenReview], [DBLP], [Semantic Scholar].
Fed-FA: Theoretically Modeling Client Data Divergence for Federated Language Backdoor Defense
Zhiyuan Zhang, Deli Chen, Hao Zhou, Fandong Meng, Jie Zhou, Xu Sun
In NeurIPS 2023
Incorporating Pre-trained Model Prompting in Multimodal Stock Volume Movement Prediction
Ruibo Chen#, Zhiyuan Zhang#, Yi Liu, Ruihan Bao, Keiko Harimoto, Xu Sun
In KDD-MLF 2023 (ML in Finance)
ASAT: Adaptively Scaled Adversarial Training in Time Series
Zhiyuan Zhang, Wei Li, Ruihan Bao, Keiko Harimoto, Yunfang Wu, Xu Sun
In Neurocomputing 2023
Diffusion Theory as a Scalpel: Detecting and Purifying Poisonous Dimensions in Pre-trained Language Models Caused by Backdoor or Bias
Zhiyuan Zhang, Deli Chen, Hao Zhou, Fandong Meng, Jie Zhou, Xu Sun
In Findings of ACL 2023
No Stock is an Island: Learning Internal and Relational Attributes of Stocks with Contrastive Learning
Shicheng Li#, Wei Li#, Zhiyuan Zhang, Ruihan Bao, Keiko Harimoto, Xu Sun
Comment: ⭐ Selected as the best paper in IJCAI-FinNLP 2022!
In IJCAI-FinNLP 2022 (Financial Technology and NLP)
How to Inject Backdoors with Better Consistency: Logit Anchoring on Clean Data
Zhiyuan Zhang, Lingjuan Lyu, Weiqiang Wang, Lichao Sun, Xu Sun
Comment: Also related to our previous paper: Neural Network Surgery [pdf].
In ICLR 2022
GA-SAM: Gradient-Strength based Adaptive Sharpness-Aware Minimization for Improved Generalization
Zhiyuan Zhang, Ruixuan Luo, Qi Su and Xu Sun
Comment: The theoretical analysis of why and how the parameter robustness relates to the generalization ability via the distributional shift between training and test sets as a bridge. Also related to our previous papers: Parameter Corruption [pdf] and Adversarial Parameter Defense [pdf].
In EMNLP 2022
PKUSEG: A Toolkit for Multi-Domain Chinese Word Segmentation
Ruixuan Luo, Jingjing Xu, Yi Zhang, Zhiyuan Zhang, Xuancheng Ren, Xu Sun
Comment: ⭐ A highly influential Chinese word segmentation open source toolkit: PKUSEG [GitHub].
Preprint
Exploring the Vulnerability of Deep Neural Networks: A Study of Parameter Corruption
Xu Sun#, Zhiyuan Zhang#, Xuancheng Ren, Ruixuan Luo, Liangyou Li
Comment: The analysis of the robustness against parameter corruptions or perturbations.
In AAAI 2021
Building an Ellipsis-aware Chinese Dependency Treebank for Web Text
Xuancheng Ren, Xu Sun, Ji Wen, Bingzhen Wei, Weidong Zhan, Zhiyuan Zhang
Comment: Releasing a Chinese dependency treebank dataset [GitHub] of 319 weibos, containing 572 sentences with omissions restored and contexts reserved.
In LREC 2018
一种模型训练方法及装置
China Patent, No.: 202110221446.7, 2022
Zhiyuan Zhang, Xu Sun, Ruixuan Luo, Bin He (Huawei)
一种模型的更新方法、装置及设备
China Patent, No.: 202111094859.X, 2021
Zhiyuan Zhang, Lingjuan Lyu, Weiqiang Wang (Ant Group)
一种模型训练方法及装置
China Patent, No.: 202110475677.0, 2021
Zhiyuan Zhang, Xuancheng Ren, Xu Sun, Bin He, Li Qian (Huawei)
Zhiyuan Zhang has a solid foundation in mathematical and programming.
He has scored 95+/100 in all the following courses:
Enhancing Byzantine-Resistant Aggregations with Client Embedding
Zhiyuan Zhang, Hao Zhou, Fandong Meng, Jie Zhou, Xu Sun
In Findings of EMNLP 2024
Fed-FA: Theoretically Modeling Client Data Divergence for Federated Language Backdoor Defense
Zhiyuan Zhang, Deli Chen, Hao Zhou, Fandong Meng, Jie Zhou, Xu Sun
In NeurIPS 2023
Diffusion Theory as a Scalpel: Detecting and Purifying Poisonous Dimensions in Pre-trained Language Models Caused by Backdoor or Bias
Zhiyuan Zhang, Deli Chen, Hao Zhou, Fandong Meng, Jie Zhou, Xu Sun
In Findings of ACL 2023
GA-SAM: Gradient-Strength based Adaptive Sharpness-Aware Minimization for Improved Generalization
Zhiyuan Zhang, Ruixuan Luo, Qi Su and Xu Sun
Comment: The theoretical analysis of why and how the parameter robustness relates to the generalization ability via the distributional shift between training and test sets as a bridge. Also related to our previous papers: Parameter Corruption [pdf] and Adversarial Parameter Defense [pdf].
In EMNLP 2022
Dim-Krum: Backdoor-Resistant Federated Learning for NLP with Dimension-wise Krum-Based Aggregation
Zhiyuan Zhang, Qi Su and Xu Sun
In Findings of EMNLP 2022
Fine-mixing: Mitigating Backdoors in Fine-tuned Language Models
Zhiyuan Zhang, Lingjuan Lyu, Xingjun Ma, Chenguang Wang and Xu Sun
In Findings of EMNLP 2022
How to Inject Backdoors with Better Consistency: Logit Anchoring on Clean Data
Zhiyuan Zhang, Lingjuan Lyu, Weiqiang Wang, Lichao Sun, Xu Sun
Comment: Also related to our previous paper: Neural Network Surgery [pdf].
In ICLR 2022
Adversarial Parameter Defense by Multi-step Risk Minimization
Zhiyuan Zhang, Ruixuan Luo, Xuancheng Ren, Qi Su, Liangyou Li, Xu Sun
Comment: The analysis of the generalization ability and the parameter robustness. A substantial journal extension of our previous conference paper: Parameter Corruption [pdf].
In Neural Networks 2021
Neural Network Surgery: Injecting Data Patterns into Pre-trained Models with Minimal Instance-wise Side Effects
Zhiyuan Zhang, Xuancheng Ren, Qi Su, Xu Sun, Bin He
Comment: Besides injecting data patterns with minimal instance-wise side effects, also about NLP backdoors for both benign and malicious purposes.
In NAACL 2021
Exploring the Vulnerability of Deep Neural Networks: A Study of Parameter Corruption
Xu Sun#, Zhiyuan Zhang#, Xuancheng Ren, Ruixuan Luo, Liangyou Li
Comment: The analysis of the robustness against parameter corruption or perturbation.
In AAAI 2021
Memorized Sparse Backpropagation
Zhiyuan Zhang, Pengcheng Yang, Xuancheng Ren, Qi Su, Xu Sun
In Neurocomputing 2020
Expose Backdoors on the Way: A Feature-Based Efficient Defense against Textual Backdoor Attacks
Sishuo Chen, Wenkai Yang, Zhiyuan Zhang, Xiaohan Bi and Xu Sun
In Findings of EMNLP 2022
Be Careful about Poisoned Word Embeddings: Exploring the Vulnerability of the Embedding Layers in NLP Models
Wenkai Yang, Lei Li, Zhiyuan Zhang, Xuancheng Ren, Xu Sun, Bin He
In NAACL 2021
Raise a Child in Large Language Model: Towards Effective and Generalizable Fine-tuning
Runxin Xu#, Fuli Luo#, Zhiyuan Zhang, Chuanqi Tan, Baobao Chang, Songfang Huang, Fei Huang
Comment: Besides the effective and generalizable fine-tuning approach, also related to the theoretical analysis of the generalization ability and the parameter robustness.
In EMNLP 2021
Understanding and Improving Layer Normalization
Jingjing Xu, Xu Sun, Zhiyuan Zhang, Guangxiang Zhao, Junyang Lin
In NeurIPS 2019
Rethinking Skip Connection with Layer Normalization
Fenglin Liu#, Xuancheng Ren#, Zhiyuan Zhang, Xu Sun, Yuexian Zou
In COLING 2020
PKUSEG: A Toolkit for Multi-Domain Chinese Word Segmentation
Ruixuan Luo, Jingjing Xu, Yi Zhang, Zhiyuan Zhang, Xuancheng Ren, Xu Sun
Comment: ⭐ A highly influential Chinese word segmentation open source toolkit: PKUSEG [GitHub].
Preprint
Pretrain-KGE: Learning Knowledge Representation from Pretrained Language Models
Zhiyuan Zhang, Xiaoqian Liu, Yi Zhang, Qi Su, Xu Sun and Bin He
Comment: The full version: preprint [OpenReview].
In Findings of EMNLP 2020 (short)
Automatic Translating between Ancient Chinese and Contemporary Chinese with Limited Aligned Corpora
Zhiyuan Zhang, Wei Li, Qi Su
In NLPCC 2019 (short)
Primal Meaning Recommendation for Chinese Expressions via Descriptions in On-line Encyclopedia
Zhiyuan Zhang, Wei Li, Jingjing Xu, Xu Sun
Preprint
Building an Ellipsis-aware Chinese Dependency Treebank for Web Text
Xuancheng Ren, Xu Sun, Ji Wen, Bingzhen Wei, Weidong Zhan, Zhiyuan Zhang
Comment: Releasing a Chinese dependency treebank dataset [GitHub] of 319 weibos, containing 572 sentences with omissions restored and contexts reserved.
In LREC 2018
Explicit Sparse Transformer: Concentrated Attention Through Explicit Selection
Guangxiang Zhao, Junyang Lin, Zhiyuan Zhang, Xuancheng Ren, Qi Su, Xu Sun
Preprint
MUSE: Parallel Multi-Scale Attention for Sequence to Sequence Learning
Guangxiang Zhao, Xu Sun, Jingjing Xu, Zhiyuan Zhang, Liangchen Luo
Preprint
No Stock is an Island: Learning Internal and Relational Attributes of Stocks with Contrastive Learning
Shicheng Li#, Wei Li#, Zhiyuan Zhang, Ruihan Bao, Keiko Harimoto, Xu Sun
Comment: ⭐ Selected as the best paper in IJCAI-FinNLP 2022!
In IJCAI-FinNLP 2022 (Financial Technology and NLP)
ASAT: Adaptively Scaled Adversarial Training in Time Series
Zhiyuan Zhang, Wei Li, Ruihan Bao, Keiko Harimoto, Yunfang Wu, Xu Sun
In Neurocomputing 2022
Incorporating Pre-trained Model Prompting in Multimodal Stock Volume Movement Prediction
Ruibo Chen#, Zhiyuan Zhang#, Yi Liu, Ruihan Bao, Keiko Harimoto, Xu Sun
In KDD-MLF 2023 (ML in Finance)
Distributional Correlation–Aware Knowledge Distillation for Stock Trading Volume Prediction
Lei Li, Zhiyuan Zhang, Ruihan Bao, Keiko Harimoto, Xu Sun
In ECML-PKDD 2022
Stock Trading Volume Prediction with Dual-Process Meta-Learning
Ruibo Chen, Wei Li, Zhiyuan Zhang, Ruihan Bao, Keiko Harimoto, Xu Sun
In ECML-PKDD 2022
Incremental Stock Volume Prediction with Gradient Distillation and Diversified Memory Selection
Shicheng Li, Zhiyuan Zhang, Ruihan Bao, Keiko Harimoto, Xu Sun
In IJCAI-AI4TS 2022 (AI for Time Series Analysis)
Learning Robust Representation for Clustering through Locality Preserving Variational Discriminative Network
Ruixuan Luo, Wei Li, Zhiyuan Zhang, Ruihan Bao, Keiko Harimoto, Xu Sun
In AAAI-RSEML 2021 (Robust, Secure, and Efficient ML)