Search

Your search keyword '"Chen, Qizhou"' showing total 29 results
Start Over Author "Chen, Qizhou"
29 results on '"Chen, Qizhou"'

1. DAFNet: Dynamic Auxiliary Fusion for Sequential Model Editing in Large Language Models

Author

Zhang, Taolin, Chen, Qizhou, Li, Dongyang, Wang, Chengyu, He, Xiaofeng, Huang, Longtao, Xue, Hui, and Huang, Jun

Subjects
Computer Science - Computation and Language
Abstract

Recently, while large language models (LLMs) have demonstrated impressive results, they still suffer from hallucination, i.e., the generation of false information. Model editing is the task of fixing factual mistakes in LLMs; yet, most previous works treat it as a one-time task, paying little attention to ever-emerging mistakes generated by LLMs. We address the task of sequential model editing (SME) that aims to rectify mistakes continuously. A Dynamic Auxiliary Fusion Network (DAFNet) is designed to enhance the semantic interaction among the factual knowledge within the entire sequence, preventing catastrophic forgetting during the editing process of multiple knowledge triples. Specifically, (1) for semantic fusion within a relation triple, we aggregate the intra-editing attention flow into auto-regressive self-attention with token-level granularity in LLMs. We further leverage multi-layer diagonal inter-editing attention flow to update the weighted representations of the entire sequence-level granularity. (2) Considering that auxiliary parameters are required to store the knowledge for sequential editing, we construct a new dataset named \textbf{DAFSet}, fulfilling recent, popular, long-tail and robust properties to enhance the generality of sequential editing. Experiments show DAFNet significantly outperforms strong baselines in single-turn and sequential editing. The usage of DAFSet also consistently improves the performance of other auxiliary network-based methods in various scenarios, Comment: ACL2024 findings

Published
2024

2. Single image super-resolution based on trainable feature matching attention network

Author

Chen, Qizhou and Shao, Qing

Subjects
Computer Science - Computer Vision and Pattern Recognition
Abstract

Convolutional Neural Networks (CNNs) have been widely employed for image Super-Resolution (SR) in recent years. Various techniques enhance SR performance by altering CNN structures or incorporating improved self-attention mechanisms. Interestingly, these advancements share a common trait. Instead of explicitly learning high-frequency details, they learn an implicit feature processing mode that utilizes weighted sums of a feature map's own elements for reconstruction, akin to convolution and non-local. In contrast, early dictionary-based approaches learn feature decompositions explicitly to match and rebuild Low-Resolution (LR) features. Building on this analysis, we introduce Trainable Feature Matching (TFM) to amalgamate this explicit feature learning into CNNs, augmenting their representation capabilities. Within TFM, trainable feature sets are integrated to explicitly learn features from training images through feature matching. Furthermore, we integrate non-local and channel attention into our proposed Trainable Feature Matching Attention Network (TFMAN) to further enhance SR performance. To alleviate the computational demands of non-local operations, we propose a streamlined variant called Same-size-divided Region-level Non-Local (SRNL). SRNL conducts non-local computations in parallel on blocks uniformly divided from the input feature map. The efficacy of TFM and SRNL is validated through ablation studies and module explorations. We employ a recurrent convolutional network as the backbone of our TFMAN to optimize parameter utilization. Comprehensive experiments on benchmark datasets demonstrate that TFMAN achieves superior results in most comparisons while using fewer parameters. The code is available at https://github.com/qizhou000/tfman., Comment: 35pages, 12 figures

Published
2024
Full Text
View/download PDF

3. Lifelong Knowledge Editing for LLMs with Retrieval-Augmented Continuous Prompt Learning

Author

Chen, Qizhou, Zhang, Taolin, He, Xiaofeng, Li, Dongyang, Wang, Chengyu, Huang, Longtao, and Xue, Hui

Subjects
Computer Science - Computation and Language
Abstract

Model editing aims to correct outdated or erroneous knowledge in large language models (LLMs) without the need for costly retraining. Lifelong model editing is the most challenging task that caters to the continuous editing requirements of LLMs. Prior works primarily focus on single or batch editing; nevertheless, these methods fall short in lifelong editing scenarios due to catastrophic knowledge forgetting and the degradation of model performance. Although retrieval-based methods alleviate these issues, they are impeded by slow and cumbersome processes of integrating the retrieved knowledge into the model. In this work, we introduce RECIPE, a RetriEval-augmented ContInuous Prompt lEarning method, to boost editing efficacy and inference efficiency in lifelong learning. RECIPE first converts knowledge statements into short and informative continuous prompts, prefixed to the LLM's input query embedding, to efficiently refine the response grounded on the knowledge. It further integrates the Knowledge Sentinel (KS) that acts as an intermediary to calculate a dynamic threshold, determining whether the retrieval repository contains relevant knowledge. Our retriever and prompt encoder are jointly trained to achieve editing properties, i.e., reliability, generality, and locality. In our experiments, RECIPE is assessed extensively across multiple LLMs and editing datasets, where it achieves superior editing performance. RECIPE also demonstrates its capability to maintain the overall performance of LLMs alongside showcasing fast editing and inference speed., Comment: 14 pages, 4 figures, 6 tables

Published
2024

4. R4: Reinforced Retriever-Reorder-Responder for Retrieval-Augmented Large Language Models

Author

Zhang, Taolin, Li, Dongyang, Chen, Qizhou, Wang, Chengyu, Huang, Longtao, Xue, Hui, He, Xiaofeng, and Huang, Jun

Subjects
Computer Science - Computation and Language
Abstract

Retrieval-augmented large language models (LLMs) leverage relevant content retrieved by information retrieval systems to generate correct responses, aiming to alleviate the hallucination problem. However, existing retriever-responder methods typically append relevant documents to the prompt of LLMs to perform text generation tasks without considering the interaction of fine-grained structural semantics between the retrieved documents and the LLMs. This issue is particularly important for accurate response generation as LLMs tend to "lose in the middle" when dealing with input prompts augmented with lengthy documents. In this work, we propose a new pipeline named "Reinforced Retriever-Reorder-Responder" (R$^4$) to learn document orderings for retrieval-augmented LLMs, thereby further enhancing their generation abilities while the large numbers of parameters of LLMs remain frozen. The reordering learning process is divided into two steps according to the quality of the generated responses: document order adjustment and document representation enhancement. Specifically, document order adjustment aims to organize retrieved document orderings into beginning, middle, and end positions based on graph attention learning, which maximizes the reinforced reward of response quality. Document representation enhancement further refines the representations of retrieved documents for responses of poor quality via document-level gradient adversarial learning. Extensive experiments demonstrate that our proposed pipeline achieves better factual question-answering performance on knowledge-intensive tasks compared to strong baselines across various public datasets. The source codes and trained models will be released upon paper acceptance., Comment: need to further experiment

Published
2024

7. Reticular Ratchets for Directing Electrochemiluminescence

Author

Luo, Rengan, primary, Luo, Xiao, additional, Xu, Haocheng, additional, Wan, Sushu, additional, Lv, Haifeng, additional, Zou, Beier, additional, Wang, Yufei, additional, Liu, Tianrui, additional, Wu, Chuang, additional, Chen, Qizhou, additional, Yu, Siqi, additional, Dong, Pengfei, additional, Tian, Yuxi, additional, Xi, Kai, additional, Yuan, Shuai, additional, Wu, Xiaojun, additional, Ju, Huangxian, additional, and Lei, Jianping, additional

Published
2024
Full Text
View/download PDF

12. Comparative removal of pharmaceuticals in aqueous phase by agricultural waste‐based biochars.

Author

Fu, Bomin, Chen, Qizhou, Sleiman, Mohamad, Ferronato, Corinne, Fine, Ludovic, Meunier, Frederic, Ferro Fernandez, Victor R., Valverde, Jose Luis, Giroir‐Fendler, Anne, Wu, Yang, Wang, Hongtao, Ma, Yibing, and Chovelon, Jean‐Marc

Subjects
*BIOCHAR, *AGRICULTURE, *DRUG adsorption, *HYDROGEN bonding interactions, *WHEAT straw, *QUANTUM wells
Abstract

The intensification of pharmaceutical use globally has led to an increase in the number of water bodies contaminated by drugs, and an effective strategy must be developed to address this issue. In this work, several biochars produced from Miscanthus straw pellets (MSP550, MSP700) and wheat straw pellets (WSP550, WSP700) at 550 and 700°C, respectively, were selected as adsorbents for removing various pharmaceuticals, such as pemetrexed (PEME), sulfaclozine (SCL), and terbutaline (TBL), from the aqueous phase. The biochar characterizations (physicochemical properties, textural properties, morphological structures, and zeta potentials) and adsorptive conditions (contact times, temperatures, and pH effect) were investigated. The infrared and Raman spectra of biochars before and after pharmaceutical adsorption, as well as quantum chemical computations, were carried out to explore the adsorption mechanisms. The results showed that the general adsorption abilities of biochars for pharmaceuticals were in the order of WSP700 > MSP700 > MSP550 > WSP550. Both the higher drug concentration and higher temperature improved biochar adsorption. By decreasing the pH, the adsorption amounts increased for PEME and SCL. However, TBL exhibited the best adsorption at pH 7, whereas a weakening of affinity occurred at lower or higher pH values. Electrostatic interactions and hydrogen bonding were the main adsorptive mechanisms between all biochars and pharmaceuticals. π–π interactions played a role in the adsorption process of low‐temperature‐prepared biochars (MSP550 and WSP550). This work can provide new insights into the control of pharmaceuticals from water with low‐cost adsorbents. Practitioner Points: Use of biochars for pharmaceuticals removal from aqueous phase.Characterization of biochars : physical and chemical properties, textural and surface properties.Simulation calculation for characterization of pharmaceuticals.Kinetic studies of pharmaceuticals adsorption on biochars.DRIFTS and Raman analysis for the understanding of adsorption process. [ABSTRACT FROM AUTHOR]

Published
2024
Full Text
View/download PDF

13. Self-Assembled Amphiphilic Chitosan Nanomicelles: Synthesis, Characterization and Antibacterial Activity.

Author

Qi, Yi, Chen, Qizhou, Cai, Xiaofen, Liu, Lifen, Jiang, Yuwei, Zhu, Xufeng, Huang, Zhicheng, Wu, Kefeng, Luo, Hui, and Ouyang, Qianqian

Subjects
*ANTIBACTERIAL agents, *SUCCINIC anhydride, *DEOXYCHOLIC acid, *DRUG carriers, *CELL membranes
Abstract

Although amphiphilic chitosan has been widely studied as a drug carrier for drug delivery, fewer studies have been conducted on the antimicrobial activity of amphiphilic chitosan. In this study, we successfully synthesized deoxycholic acid-modified chitosan (CS-DA) by grafting deoxycholic acid (DA) onto chitosan C2-NH2, followed by grafting succinic anhydride, to prepare a novel amphiphilic chitosan (CS-DA-SA). The substitution degree was 23.93% for deoxycholic acid and 29.25% for succinic anhydride. Both CS-DA and CS-DA-SA showed good blood compatibility. Notably, the synthesized CS-DA-SA can self-assemble to form nanomicelles at low concentrations in an aqueous environment. The results of CS, CS-DA, and CS-DA-SA against Escherichia coli and Staphylococcus aureus showed that CS-DA and CS-DA-SA exhibited stronger antimicrobial effects than CS. CS-DA-SA may exert its antimicrobial effect by disrupting cell membranes or forming a membrane on the cell surface. Overall, the novel CS-DA-SA biomaterials have a promising future in antibacterial therapy. [ABSTRACT FROM AUTHOR]

Published
2023
Full Text
View/download PDF

14. Time-dependent simulation of photocurrent-detected two-dimensional spectroscopy of open systems.

Author

Chen, Qizhou, Kwok, Yan Ho, Zhou, Weijun, Chen, Guanhua, and Mukamel, Shaul

Subjects
*TIME-dependent density functional theory, *QUANTUM dots spectra, *ELECTRONIC spectra, *LASER pulses, *SPECTROMETRY, *PHOTOELECTRIC cells
Abstract

A new time-domain simulation protocol of two-dimensional electronic spectra with photocurrent detection is presented. Time-dependent density functional theory for open systems at finite temperature is applied to evaluate the photocurrent response to four laser pulses, and a non-perturbative phase-matching approach is implemented to extract the fourth-order photocurrent signal with a desired phase-matching condition. Simulations for an open three-level model indicates that transition dipoles interact resonantly with the incident pulses and that different sample–electrode couplings may be identified by appearance of different peaks/valleys in photocurrent spectra from different electrodes. Moreover, qualitative reproduction of experimental spectra of a PbS quantum dot photocell [Karki et al., Nat. Commun. 5(1), 5869 (2014)] reveals the stimulated electron dynamics. [ABSTRACT FROM AUTHOR]

Published
2021
Full Text
View/download PDF

20. Recovery of Fe, Ni, Co, and Cu from Nickel Converter Slag through Oxidation and Reduction

Author

Chen Qizhou, Hu Shaoyan, Guo Yaguang, Guangsheng Wei, Rong Zhu, and Wang Yun

Subjects
Materials science, Mechanical Engineering, Metallurgy, Metals and Alloys, Slag, chemistry.chemical_element, 02 engineering and technology, 021001 nanoscience & nanotechnology, Redox, 020501 mining & metallurgy, Flue-gas desulfurization, Reduction (complexity), Nickel, 0205 materials engineering, chemistry, Mechanics of Materials, visual_art, Materials Chemistry, visual_art.visual_art_medium, 0210 nano-technology
Published
2018

21. Expérimentations et simulations sur des micelles wormlike : comportement photo-réactif et applications

Author

Chen, Qizhou and STAR, ABES

Subjects
Bubble rise, [CHIM.THEO] Chemical Sciences/Theoretical and/or physical chemistry, Viscoelastic fluid, Ascension de la bulle, Molecular dynamics simulation, [PHYS.PHYS.PHYS-COMP-PH] Physics [physics]/Physics [physics]/Computational Physics [physics.comp-ph], Simulation de dynamique moléculaire, Fluide viscoélastique, [PHYS.PHYS.PHYS-FLU-DYN] Physics [physics]/Physics [physics]/Fluid Dynamics [physics.flu-dyn], Micelle en forme de ver photosensible, Photo-responsive wormlike micelle, [PHYS.COND.CM-SCM] Physics [physics]/Condensed Matter [cond-mat]/Soft Condensed Matter [cond-mat.soft]
Abstract

Responsive soft materials have been an active research topic for decades. Among the different types of materials photo-responsive wormlike micelle systems constructed by surfactants with controllable self-assembly morphology and adjustable rheological properties have attracted widespread attention. Such photo-responsive self-assembly systems can not only be easy to design, clean and pollution-free, but it is also possible to precisely adjust and control the response time and localization of the stimulus. This is why such systems show promising application prospects in drug release, petrochemical industry, switchable catalysis, surface modification, bioengineering, smart photosensitive materials and optics field. One such system is composed of cinnamate-based photosensitive molecules and Gemini surfactants. It shows a rich zoology of self-assembly morphologies, including spherical micelles, wormlike micelles, vesicles, liquid crystals, an aqueous two-phase system and so on. These self-assembly structures all show excellent photo-responsive behavior. However, owing to the microscopic scale of the self-assemblies, the question of how they assemble and what drives the morphology change under UV irradiation is still open. With the rapid improvement of computer performance and the progress of theoretical methods, computer simulation and calculation methods have become an indispensable method in theoretical research. In this thesis, to construct the photo-responsive wormlike micelles, the cinnamates, represented by o-methoxycinnamate (OMCA), with different ortho-substituents in meta- or para-position were select as the photosensitive molecules to introduce into Gemini surfactant solution. Through experimental methods such as rheology and reaction kinetics, combined with molecular dynamics (MD) simulation, quantitative calculations, the morphological transformation process under the UV light, photo-isomerization reaction kinetics, the difference in photosensitive molecules and the mechanism of morphology transformation of this system are comprehensively investigated. The rheology of wormlike micelles makes them an interesting shear thinning model fluid. The collective behavior of bubbles in shear thinning fluids is common in industrial processes such as sewage treatment, oil extraction, and froth flotation. Because of the complex rheological behavior of viscoelastic fluids, the research on the rising behavior of bubbles in the fluid is scarce. We studied the rise of bubbles in a classic wormlike micelle system (composed of CTAB/NaSal) in bubbly liquids with a gas fraction of 5%-30% and show that the rise of the concentrated suspensions of bubbles in the Newtonian fluids is well described using the Richardson-Zaki equation. However, for the shear thinning solutions, the Richardson-Zaki relation no longer applies. Bubble clustering leads to faster rise velocities and a weaker dependence on the bubble volume fraction. We discuss these results in terms of the flotation process, where they can help design more efficient flotation experiments for shear thinning fluids. Our research will help understanding of the influence of the structure of photo-sensitive molecules on the photo-responsive behavior of photo-responsive wormlike micelles as well as the bubbles rise in non-Newtonian fluid, which further provides theoretical guidance for industrial applications such as photo-rheological fluid, gas-liquid separation, foam flotation, etc., Les matériaux souples réactifs sont un sujet de recherche actif depuis des décennies. Parmi les différents types de matériaux, les systèmes de micelles ressemblant à des vers photosensibles construits par des tensioactifs avec une morphologie d'auto-assemblage contrôlable et des propriétés rhéologiques réglables ont attiré une large attention. De tels systèmes d'auto-assemblage photosensibles peuvent non seulement être faciles à concevoir, propres et sans pollution, mais il est également possible d'ajuster et de contrôler avec précision le temps de réponse et la localisation du stimulus. L'un de ces systèmes est composé de molécules photosensibles à base de cinnamate et de tensioactifs Gemini. Il montre une riche zoologie de morphologies d'auto-assemblage, y compris des micelles sphériques, des micelles vermiformes, des vésicules, des cristaux liquides, un système aqueux à deux phases, etc. Ces auto-assemblages présentent tous un excellent comportement photosensible. Cependant, en raison de l'échelle microscopique des auto-assemblages, la question de savoir comment ils s'assemblent et ce qui entraîne le changement de morphologie sous irradiation UV est encore ouverte. Les méthodes de simulation informatique à l'échelle atomique et moléculaire suscitent notre intérêt en raison du succès de la modélisation du comportement d'auto-assemblage. Avec l'amélioration rapide des performances des ordinateurs et les progrès des méthodes théoriques, les méthodes de simulation et de calcul sur ordinateur sont devenues une méthode indispensable dans la recherche théorique. Dans cette thèse, pour construire les micelles wormlike photosensibles, les cinnamates, représentés par l'o-méthoxycinnamate, avec différents ortho-substituants en position méta-, para ont été sélectionnés comme molécules photosensiblesà introduire dans la solution de tensioactif Gemini. Grâce à des méthodes expérimentales telles que la rhéologie et la cinétique de réaction, combinées à une simulation de dynamique moléculaire, des calculs quantitatifs, le processus de transformation morphologique sous la lumière UV, la cinétique de réaction de photo-isomérisation, la différence de molécules photosensibles et le mécanisme de transformation morphologique de cette système font l'objet d'une enquête approfondie. La rhéologie des micelles vermiformes en fait un fluide modèle intéressant pour l'amincissement par cisaillement. Le comportement collectif des bulles dans les fluides fluidifiants par cisaillement est courant dans les procédés industriels tels que le traitement des eaux usées, l'extraction du pétrole et la flottation par mousse. En raison du comportement rhéologique complexe des fluides viscoélastiques, les recherches sur le comportement ascendant des bulles dans le fluide sont rares. Nous avons étudié la montée de bulles dans un système micellaire vermiforme classique (composé de CTAB/NaSal) dans des liquides bulleux avec une fraction gazeuse de 5% à 30% et montrons que la montée des suspensions concentrées de bulles dans les fluides newtoniens est bien décrite en utilisant l'équation de Richardson-Zaki. Cependant, pour les solutions d'amincissement par cisaillement, la relation de Richardson-Zaki ne s'applique plus. Le regroupement des bulles entraîne des vitesses de montée plus rapides et une dépendance plus faible de la fraction volumique des bulles. Nous discutons de ces résultats en termes de processus de flottation, où ils peuvent aider à concevoir des expériences de flottation plus efficaces pour les fluides fluidifiants par cisaillement. Notre recherche aidera à comprendre l'influence de la structure des molécules photosensibles sur le comportement photosensible des micelles de type ver photosensibles ainsi que la montée des bulles dans le fluide non newtonien, ce qui fournit en outre des orientations théoriques pour les applications industrielles telles que fluide photo-rhéologique, séparation gaz-liquide, flottation de mousse etc.

Published
2021

23. Experimental Study on Oxidative Desulfurization and Selective Reduction of Molten Copper Slag

Author

Chen Qizhou, Wang Yun, and Zhu Rong

Subjects
Materials science, Metallurgy, chemistry.chemical_element, Copper, Sulfur, Redox, Copper slag, Flue-gas desulfurization, Metal, Recovery rate, chemistry, visual_art, visual_art.visual_art_medium, Selective reduction
Abstract

The copper in slag is mainly in the form of Cu2S, and it’s easy to make matte suspended in slag. The sulfur in slag can be removed by oxidative process, and the copper sulfides can be converted into oxides, then the copper can be easily reduced and separated. Thermodynamic calculation by FactSage7.0 shows that the sulfur in slag can be removed by air, while the activity of Cu2O will increase, and the Cu2S cannot be reduced, while Cu2O is easily reduced into metal. The oxidation and reduction of molten copper slag were carried out and agreed well with theoretical calculations. Sulfur in slag can be dropped to 0.23 wt%, and the utilization rate of air is in the range of 14.98–73.39%. The copper recovery rate can reach 83.82% and the residue copper can be reduced to 0.28 wt%.

Published
2018

Catalog

Books, media, physical & digital resources
See catalog results