Qiang Wei
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Qiang Wei has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
Guoqing Chen
9 shared publications- Guoqing Chen
School of Economics and Management, Tsinghua University, Beijing, China
9 shared publications Xiaowei Shi
5 shared publicationsCong Wang
3 shared publications- Dandan Qiao
Department of Information Systems and Analytics, National University of Singapore, Singapore 117417, Singapore
3 shared publications - Xunhua Guo
Tsinghua University
3 shared publications
Latest Publications
- 11 May 2026
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Management Science
Regurgitative Training: The Value of Real Data in Training Large Language Models
What happens if we train a new large language model (LLM) using data at least partially generated by other LLMs? The explosive success of LLMs means that content online will increasingly be generated by LLMs rather than humans, which inevitably enters the training data sets of next-generation LLMs. In this paper, we study the implications of such “regurgitative training” on LLM performance. Starting with the machine translation task (a representative language task with well-established evaluation criteria), we fine-tune LLMs with data generated either by themselves or by other LLMs, and we find strong evidence that regurgitative training handicaps the performance of fine-tuned LLMs. A comparison between LLM-generated data and real data reveals suggestive evidence that higher error rates and lower lexical diversity in LLM-generated data may be at play. Accordingly, we propose and evaluate three strategies to mitigate the performance loss by (i) prioritizing high-quality LLM-generated data, (ii) mixing data generated by multiple LLMs, and (iii) prioritizing LLM-generated data that most resemble real data. All three strategies can improve the performance of regurgitative training to some extent but cannot fully close the gap from training with real data. This highlights that real, human-generated data cannot be easily substituted by LLM-generated data in training LLMs. Additionally, we investigate regurgitative training on a creative ideation task with human judgement-based evaluations. Interestingly, we find that preference-based fine-tuning with human feedback on LLM-generated ideas can actually improve ideation performance. This showcases that human preference data when combined with LLM-generated data can bring performance gains. This paper was accepted by Hemant Bhargava, information systems. Funding: This work was supported by the National Natural Science Foundation of China [Grants 72421001 and 72172070] and the Singapore Ministry of Education Academic Research Fund Tier 2 A-8003504 [Robert Brown Promising Researcher Award MOE-T2EP40]. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2024.07005 .
- 11 May 2026
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Management Science
Regurgitative Training: The Value of Real Data in Training Large Language Models
What happens if we train a new large language model (LLM) using data at least partially generated by other LLMs? The explosive success of LLMs means that content online will increasingly be generated by LLMs rather than humans, which inevitably enters the training data sets of next-generation LLMs. In this paper, we study the implications of such “regurgitative training” on LLM performance. Starting with the machine translation task (a representative language task with well-established evaluation criteria), we fine-tune LLMs with data generated either by themselves or by other LLMs, and we find strong evidence that regurgitative training handicaps the performance of fine-tuned LLMs. A comparison between LLM-generated data and real data reveals suggestive evidence that higher error rates and lower lexical diversity in LLM-generated data may be at play. Accordingly, we propose and evaluate three strategies to mitigate the performance loss by (i) prioritizing high-quality LLM-generated data, (ii) mixing data generated by multiple LLMs, and (iii) prioritizing LLM-generated data that most resemble real data. All three strategies can improve the performance of regurgitative training to some extent but cannot fully close the gap from training with real data. This highlights that real, human-generated data cannot be easily substituted by LLM-generated data in training LLMs. Additionally, we investigate regurgitative training on a creative ideation task with human judgement-based evaluations. Interestingly, we find that preference-based fine-tuning with human feedback on LLM-generated ideas can actually improve ideation performance. This showcases that human preference data when combined with LLM-generated data can bring performance gains. This paper was accepted by Hemant Bhargava, information systems. Funding: This work was supported by the National Natural Science Foundation of China [Grants 72421001 and 72172070] and the Singapore Ministry of Education Academic Research Fund Tier 2 A-8003504 [Robert Brown Promising Researcher Award MOE-T2EP40]. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2024.07005 .
- 16 November 2025
- Published by Elsevier in Decision Support Systems
- Vol. 200
ProMatch: A novel dynamic process-unpacking approach for two-way proactive recruitment
- 16 November 2025
- Published by Elsevier in Decision Support Systems
- Vol. 200
ProMatch: A novel dynamic process-unpacking approach for two-way proactive recruitment
- 18 October 2025
- Published by Elsevier in Information & Management
- Vol. 63 (1)
The gap matters: An explainable customer dissatisfaction tracing analysis
- 18 October 2025
- Published by Elsevier in Information & Management
- Vol. 63 (1)
The gap matters: An explainable customer dissatisfaction tracing analysis
- 30 September 2024
- Published by Elsevier in Information & Management
- Vol. 61 (8)
Encoding consumer interests into product snippets with a multi-criteria genetic optimization approach
- 30 September 2024
- Published by Elsevier in Information & Management
- Vol. 61 (8)
Encoding consumer interests into product snippets with a multi-criteria genetic optimization approach
- 1 September 2024
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Information Systems Research
- Vol. 35 (3) , 1382-1402
Dynamic Bayesian Network–Based Product Recommendation Considering Consumers’ Multistage Shopping Journeys: A Marketing Funnel Perspective
Recommender systems are widely used by platforms/merchants to find the products that are likely to interest consumers. However, existing dynamic methods still face challenges with regard to diverse behaviors, variability in interest shifts, and the identification of psychological dynamics. Premised on the marketing funnel perspective to analyze consumer shopping journeys, this study proposes a novel and effective machine learning approach for product recommendation, namely, multi-stage dynamic Bayesian network (MS-DBN), which models the generative processes of consumers’ interactive behaviors with products in light of their stage transitions and interest shifts. In this way, consumers’ stage-interest-behavior dynamics can be learnt, especially the variability in interest shifts. This provides managerial implications for practice. MS-DBN demonstrates significant performance advantage with general applicability by extracting the generalizable regularity during shopping journeys, which compensates the diversity and sparsity frequently observed in consumer behaviors. In addition, aided by the identification strategies integrated into the learning process, the latent variables in the model can be detected such that consumers’ invisible psychological stages and interests in products can be identified from their observed behaviors, shedding light on the targeted marketing of platforms/merchants and thus enriching the practical value of the approach.
- 1 September 2024
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Information Systems Research
- Vol. 35 (3) , 1382-1402
Dynamic Bayesian Network–Based Product Recommendation Considering Consumers’ Multistage Shopping Journeys: A Marketing Funnel Perspective
Recommender systems are widely used by platforms/merchants to find the products that are likely to interest consumers. However, existing dynamic methods still face challenges with regard to diverse behaviors, variability in interest shifts, and the identification of psychological dynamics. Premised on the marketing funnel perspective to analyze consumer shopping journeys, this study proposes a novel and effective machine learning approach for product recommendation, namely, multi-stage dynamic Bayesian network (MS-DBN), which models the generative processes of consumers’ interactive behaviors with products in light of their stage transitions and interest shifts. In this way, consumers’ stage-interest-behavior dynamics can be learnt, especially the variability in interest shifts. This provides managerial implications for practice. MS-DBN demonstrates significant performance advantage with general applicability by extracting the generalizable regularity during shopping journeys, which compensates the diversity and sparsity frequently observed in consumer behaviors. In addition, aided by the identification strategies integrated into the learning process, the latent variables in the model can be detected such that consumers’ invisible psychological stages and interests in products can be identified from their observed behaviors, shedding light on the targeted marketing of platforms/merchants and thus enriching the practical value of the approach.
- 7 February 2024
- Published by Springer Nature in Frontiers of Engineering Management
- Vol. 11 (1) , 128-142
A bilateral heterogeneous graph model for interpretable job recommendation considering both reciprocity and competition
- 7 February 2024
- Published by Springer Nature in Frontiers of Engineering Management
- Vol. 11 (1) , 128-142
A bilateral heterogeneous graph model for interpretable job recommendation considering both reciprocity and competition
- 1 December 2023
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Information Systems Research
- Vol. 34 (4) , 1641-1663
When Images Backfire: The Effect of Customer-Generated Images on Product Rating Dynamics
Customer-generated images (CGIs) are images posted by customers on e-commerce platforms, and they usually appear in the review sections together with review text and ratings provided by customers having purchase experiences. Despite their prevalent adoption by e-commerce platforms, the effect of CGIs on customers’ postpurchase satisfaction remains unclear. We find that CGIs lead to a decline in subsequent ratings compared with product ratings not exposed to CGIs. Furthermore, high CGI review ratings and high aesthetic quality exacerbate the negative effect, whereas reviewers’ face disclosure in CGIs can alleviate the negative effect. Through cross-product analyses, we find that the negative effect is more prominent for experience goods (e.g., women’s dresses) than for search goods (e.g., lightning cables). Results from a laboratory experiment show that participants experience significantly higher expectation and negative disconfirmation when reading CGI reviews with high ratings, whereas the uncertainty reduction effect is insignificant, which collectively explains the decline of subsequent product ratings from a theoretical perspective. These findings suggest that platforms and retailers should be aware of the potential negative effect of CGIs on the rating dynamics and take appropriate measures to circumvent it.
- 1 December 2023
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Information Systems Research
- Vol. 34 (4) , 1641-1663
When Images Backfire: The Effect of Customer-Generated Images on Product Rating Dynamics
Customer-generated images (CGIs) are images posted by customers on e-commerce platforms, and they usually appear in the review sections together with review text and ratings provided by customers having purchase experiences. Despite their prevalent adoption by e-commerce platforms, the effect of CGIs on customers’ postpurchase satisfaction remains unclear. We find that CGIs lead to a decline in subsequent ratings compared with product ratings not exposed to CGIs. Furthermore, high CGI review ratings and high aesthetic quality exacerbate the negative effect, whereas reviewers’ face disclosure in CGIs can alleviate the negative effect. Through cross-product analyses, we find that the negative effect is more prominent for experience goods (e.g., women’s dresses) than for search goods (e.g., lightning cables). Results from a laboratory experiment show that participants experience significantly higher expectation and negative disconfirmation when reading CGI reviews with high ratings, whereas the uncertainty reduction effect is insignificant, which collectively explains the decline of subsequent product ratings from a theoretical perspective. These findings suggest that platforms and retailers should be aware of the potential negative effect of CGIs on the rating dynamics and take appropriate measures to circumvent it.
- Open Access
- 1 September 2021
- Published by KEAI Publishing in Fundamental Research
- Vol. 1 (5) , 504-507
Big data driven management and decision sciences: A NSFC grand research plan
- 1 September 2021
- Published by KEAI Publishing in Fundamental Research
- Vol. 1 (5) , 504-507
Open AccessBig data driven management and decision sciences: A NSFC grand research plan
- 2 January 2021
- Published by Taylor & Francis in Journal of Management Information Systems
- Vol. 38 (1) , 82-107
Mitigating the Adverse Effect of Monetary Incentives on Voluntary Contributions Online
Numerous online information systems (e.g., question and answer [Q&A] forums, citizen science communities, and review websites) rely heavily on volunteer contributions. Managers have used monetary incentives to induce individuals to increase their contribution level. However, monetary incentives could also generate adverse effects, which could dampen individuals’ intrinsic motivation and lead to lower contribution quality when incentives are small. To address this issue, we propose two intervention strategies, goal-setting and challenge-seeking, and conduct a series of randomized experiments. We find that small monetary incentives, when combined with the appropriate intervention strategies, can motivate users to increase contribution quantity while simultaneously sustaining high quality. Thus, integrating such intervention strategies with small incentives can be a cost-effective way to encourage voluntary contributions. Our research contributes to the literature on incentive provision and provides theoretical and practical implications for platforms relying on voluntary contributions.
- 2 January 2021
- Published by Taylor & Francis in Journal of Management Information Systems
- Vol. 38 (1) , 82-107
Mitigating the Adverse Effect of Monetary Incentives on Voluntary Contributions Online
Numerous online information systems (e.g., question and answer [Q&A] forums, citizen science communities, and review websites) rely heavily on volunteer contributions. Managers have used monetary incentives to induce individuals to increase their contribution level. However, monetary incentives could also generate adverse effects, which could dampen individuals’ intrinsic motivation and lead to lower contribution quality when incentives are small. To address this issue, we propose two intervention strategies, goal-setting and challenge-seeking, and conduct a series of randomized experiments. We find that small monetary incentives, when combined with the appropriate intervention strategies, can motivate users to increase contribution quantity while simultaneously sustaining high quality. Thus, integrating such intervention strategies with small incentives can be a cost-effective way to encourage voluntary contributions. Our research contributes to the literature on incentive provision and provides theoretical and practical implications for platforms relying on voluntary contributions.