Zimian Wei
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Zimian Wei has a total of 22 co-authors. Here are the top co-authors by number of shared publications:
- Dongsheng Li
PDL Lab, College of Computer Science and Technology, National University of Defense Technology, Changsha, Hunan, China
2 shared publications - Hengyue Pan
National University of Defense Technology
2 shared publications - Huadong Dai
Artificial Intelligence Research Center, National Innovation Institute of Defense Technology, Beijing, China
2 shared publications - Peijie Dong
The Hong Kong University of Science and Technology (Guang Zhou), Guangzhou, China
2 shared publications - Wenjing Yang
National university of defense technology, Changsha, China
2 shared publications - Xin Niu
National University of Defense Technology
2 shared publications
Latest Publications
- Open Access
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1887-1898
Rethinking the Hidden Risk of Reranking: Achieving Risk-aware Reranking with Information Gain for RAG with LLMs
Retrieval-augmented generation (RAG) has become a cornerstone for enhancing large language models (LLMs) with real-time information from the Web, but its performance often heavily depends on the quality of the retrieved documents. Given that RAG systems frequently draw from vast and often noisy Web corpora, ensuring the reliability of retrieved content is paramount. While rerankers improve the factual accuracy of the RAG system by elevating the proportion of ground-truth documents (GD) in high-ranked results, the shifts of document type distributions during reranking remain unclear, hindering the understanding of the reranker's behavior. To bridge this gap, we conduct an empirical study to categorize documents and compare their distribution before and after reranking. We reveal a counterintuitive finding: though rerankers improve the proportion of GD, they also significantly increase the proportion of harmful documents (HD) in top-ranked retrieved documents. It not only narrows the potential context window for ranking the GD higher but also increases the risk of HD misleading the LLMs, potentially leading to the generation and propagation of misinformation across Web platforms. Motivated by this finding, we propose a risk-aware reranking method for RAG with LLMs, which balances the risk and benefit during reranking. Given a query, the RAG framework first retrieves relevant documents. Then, our approach quantifies the potential beneficial and harmful impacts of various documents on the LLMs' generation. To estimate the impacts, we conduct a dual-aspect document impact assessment via information gain, which employs a risk clipping to avoid the numerical fluctuations in the estimation. Finally, we conduct the reranking according to the potential impact of each document, enabling the reranker to significantly reduce the HD proportion. Experiments and analysis across multiple models and datasets, including Wikipedia, web news, and research papers, show the effectiveness of our method. Our code is available at https://github.com/lzz335/hidden_risk_of_reranking.
- 12 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 1887-1898
Open AccessRethinking the Hidden Risk of Reranking: Achieving Risk-aware Reranking with Information Gain for RAG with LLMs
Retrieval-augmented generation (RAG) has become a cornerstone for enhancing large language models (LLMs) with real-time information from the Web, but its performance often heavily depends on the quality of the retrieved documents. Given that RAG systems frequently draw from vast and often noisy Web corpora, ensuring the reliability of retrieved content is paramount. While rerankers improve the factual accuracy of the RAG system by elevating the proportion of ground-truth documents (GD) in high-ranked results, the shifts of document type distributions during reranking remain unclear, hindering the understanding of the reranker's behavior. To bridge this gap, we conduct an empirical study to categorize documents and compare their distribution before and after reranking. We reveal a counterintuitive finding: though rerankers improve the proportion of GD, they also significantly increase the proportion of harmful documents (HD) in top-ranked retrieved documents. It not only narrows the potential context window for ranking the GD higher but also increases the risk of HD misleading the LLMs, potentially leading to the generation and propagation of misinformation across Web platforms. Motivated by this finding, we propose a risk-aware reranking method for RAG with LLMs, which balances the risk and benefit during reranking. Given a query, the RAG framework first retrieves relevant documents. Then, our approach quantifies the potential beneficial and harmful impacts of various documents on the LLMs' generation. To estimate the impacts, we conduct a dual-aspect document impact assessment via information gain, which employs a risk clipping to avoid the numerical fluctuations in the estimation. Finally, we conduct the reranking according to the potential impact of each document, enabling the reranker to significantly reduce the HD proportion. Experiments and analysis across multiple models and datasets, including Wikipedia, web news, and research papers, show the effectiveness of our method. Our code is available at https://github.com/lzz335/hidden_risk_of_reranking.
- Open Access
- 1 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 17030-17040
EMQ: Evolving Training-free Proxies for Automated Mixed Precision Quantization
Mixed-Precision Quantization (MQ) can achieve a competitive accuracy-complexity trade-off for models. Conventional training-based search methods require time-consuming candidate training to search optimized per-layer bit-width configurations in MQ. Recently, some training-free approaches have presented various MQ proxies and significantly improve search efficiency. However, the correlation between these proxies and quantization accuracy is poorly understood. To address the gap, we first build the MQ-Bench-101, which involves different bit configurations and quantization results. Then, we observe that the existing training-free proxies perform weak correlations on the MQ-Bench-101. To efficiently seek superior proxies, we develop an automatic search of proxies framework for MQ via evolving algorithms. In particular, we devise an elaborate search space involving the existing proxies and perform an evolution search to discover the best correlated MQ proxy. We proposed a diversity-prompting selection strategy and compatibility screening protocol to avoid premature convergence and improve search efficiency. In this way, our Evolving proxies for Mixed-precision Quantization (EMQ) framework allows the auto-generation of proxies without heavy tuning and expert knowledge. Extensive experiments on ImageNet with various ResNet and MobileNet families demonstrate that our EMQ obtains superior performance than state-of-the-art mixed-precision methods at a significantly reduced cost. The code will be released.
- 1 October 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 17030-17040
Open AccessEMQ: Evolving Training-free Proxies for Automated Mixed Precision Quantization
Mixed-Precision Quantization (MQ) can achieve a competitive accuracy-complexity trade-off for models. Conventional training-based search methods require time-consuming candidate training to search optimized per-layer bit-width configurations in MQ. Recently, some training-free approaches have presented various MQ proxies and significantly improve search efficiency. However, the correlation between these proxies and quantization accuracy is poorly understood. To address the gap, we first build the MQ-Bench-101, which involves different bit configurations and quantization results. Then, we observe that the existing training-free proxies perform weak correlations on the MQ-Bench-101. To efficiently seek superior proxies, we develop an automatic search of proxies framework for MQ via evolving algorithms. In particular, we devise an elaborate search space involving the existing proxies and perform an evolution search to discover the best correlated MQ proxy. We proposed a diversity-prompting selection strategy and compatibility screening protocol to avoid premature convergence and improve search efficiency. In this way, our Evolving proxies for Mixed-precision Quantization (EMQ) framework allows the auto-generation of proxies without heavy tuning and expert knowledge. Extensive experiments on ImageNet with various ResNet and MobileNet families demonstrate that our EMQ obtains superior performance than state-of-the-art mixed-precision methods at a significantly reduced cost. The code will be released.
- 1 January 2022
- Published by Springer Nature
- p. 875-887
Influence of Initialization and Modularization on the Performance of Network Morphism-Based Neural Architecture Search
- 1 January 2022
- Published by Springer Nature
- p. 875-887
Influence of Initialization and Modularization on the Performance of Network Morphism-Based Neural Architecture Search
- Open Access
- 13 September 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 132438-132447
Person Re-Identification Based on Multi-Parts of Local Feature Network
Person re-identification has become increasing popular because of its widely application in computer vision. In this paper, we propose a novel, simple and efficient person re-id network called MPLFN. The network combines two tasks: the classification task and the metric learning task. In the classification task, we uniformly partition N feature parts from an image, and compute the person classification loss in each part separately. Computing the part loss separately guides the network to focus on every body part and learn discriminative representations for each of them. And then in the metric learning task, we recalculate the distance of two images by the shortest path between two sets of feature parts. Then the distances are put into a triplet loss to perform a dynamic part alignment during the training. With the joint learning of these two tasks, the performance of the network is significantly enhanced. Compared with existing person re-id works, MPLFN achieves a better performance on three mainstream person re-identification datasets. Extensive experiments have been conducted to validate our proposed method.
- 13 September 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 132438-132447
Open AccessPerson Re-Identification Based on Multi-Parts of Local Feature Network
Person re-identification has become increasing popular because of its widely application in computer vision. In this paper, we propose a novel, simple and efficient person re-id network called MPLFN. The network combines two tasks: the classification task and the metric learning task. In the classification task, we uniformly partition N feature parts from an image, and compute the person classification loss in each part separately. Computing the part loss separately guides the network to focus on every body part and learn discriminative representations for each of them. And then in the metric learning task, we recalculate the distance of two images by the shortest path between two sets of feature parts. Then the distances are put into a triplet loss to perform a dynamic part alignment during the training. With the joint learning of these two tasks, the performance of the network is significantly enhanced. Compared with existing person re-id works, MPLFN achieves a better performance on three mainstream person re-identification datasets. Extensive experiments have been conducted to validate our proposed method.
- 1 March 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 254-259
A research on metric learning in computer vision and pattern recognition
Metric learning refers to the task of learning a distance function among objects. Recently, metric learning has drawn wide attention in the field of computer vision and pattern recognition. This paper seeks to examine metric learning from two different perspectives: conventional metric learning and deep metric learning. The formulation and principal of different metric learning are delineated and the advantages and disadvantages are then presented. The paper concludes with the discussion on the future direction of the development of metric learning.
- 1 March 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 254-259
A research on metric learning in computer vision and pattern recognition
Metric learning refers to the task of learning a distance function among objects. Recently, metric learning has drawn wide attention in the field of computer vision and pattern recognition. This paper seeks to examine metric learning from two different perspectives: conventional metric learning and deep metric learning. The formulation and principal of different metric learning are delineated and the advantages and disadvantages are then presented. The paper concludes with the discussion on the future direction of the development of metric learning.