Shuai Jiang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Shuai Jiang has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Hui Xiong
The Hong Kong University of Science and Technology (Guangzhou), Guangzhou, China and The Hong Kong University of Science and Technology, Hong Kong SAR, China
4 shared publications - Chuanren Liu
The University of Tennessee, Knoxville, USA
2 shared publications Yanhong Guo
2 shared publications- Yanhong Guo
School of Economics and Management, Dalian University of Technology, Dalian, China
2 shared publications - Bingzhuo Zhong
The Thrust of Artificial Intelligence, Information Hub, The Hong Kong University of Science and Technology(Guangzhou), Guangzhou, China and The Thrust of Intelligent Transportation, System Hub, The Hong Kong University of Science and Technology(Guangzhou), Guangzhou, China
1 shared publication Chuanren Liu
1 shared publication
Latest Publications
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 8753-8764
SafeBuild-Bench: A Temporal-Robust Construction Safety Benchmark with Graph-Enhanced Data Mining
Construction-safety models must handle concrete deployment risks, such as a worker standing near a scaffold edge without guardrails, rather than only recognize common objects in curated images. Yet real inspection archives are redundant, long-tailed, and collected across changing sites and months. We introduce SafeBuild-Bench, a metadata-driven benchmark for evaluating multimodal large language models on construction safety under realistic temporal and site variation. It is mined from 100K+ industrial image-text records and contains 3,314 task instances from over 3,000 expert-verified images, covering multiple-choice hazard identification and free-form hazard description. To make expert verification scalable, we develop GEMS, a graph-enhanced multimodal selection pipeline that combines a proxy-model confusion signal with graph-based diversity to identify informative candidates from redundant streams. On public instruction-tuning data, GEMS-selected subsets preserve robustness-oriented performance under small data budgets. On SafeBuild-Bench, current MLLMs remain far from reliable construction-safety understanding, with the best overall score near 60. We release the benchmark, evaluation scripts, and GEMS codebase at https://github.com/safebuild/gems.
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 8753-8764
Open AccessSafeBuild-Bench: A Temporal-Robust Construction Safety Benchmark with Graph-Enhanced Data Mining
Construction-safety models must handle concrete deployment risks, such as a worker standing near a scaffold edge without guardrails, rather than only recognize common objects in curated images. Yet real inspection archives are redundant, long-tailed, and collected across changing sites and months. We introduce SafeBuild-Bench, a metadata-driven benchmark for evaluating multimodal large language models on construction safety under realistic temporal and site variation. It is mined from 100K+ industrial image-text records and contains 3,314 task instances from over 3,000 expert-verified images, covering multiple-choice hazard identification and free-form hazard description. To make expert verification scalable, we develop GEMS, a graph-enhanced multimodal selection pipeline that combines a proxy-model confusion signal with graph-based diversity to identify informative candidates from redundant streams. On public instruction-tuning data, GEMS-selected subsets preserve robustness-oriented performance under small data budgets. On SafeBuild-Bench, current MLLMs remain far from reliable construction-safety understanding, with the best overall score near 60. We release the benchmark, evaluation scripts, and GEMS codebase at https://github.com/safebuild/gems.
- 2 December 2025
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in INFORMS Journal on Computing
Code and Data Repository for Which to Trust: Extracting Collective Wisdom Based on Opinion Quality Rank Learning
This repository contains the implementation of a novel deep learning model, CrowdRank, for integrating heterogeneous expert opinions, as presented in our paper "Which to Trust: Extracting Collective Wisdom Based on Opinion Quality Rank Learning".
- 2 December 2025
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in INFORMS Journal on Computing
Code and Data Repository for Which to Trust: Extracting Collective Wisdom Based on Opinion Quality Rank Learning
This repository contains the implementation of a novel deep learning model, CrowdRank, for integrating heterogeneous expert opinions, as presented in our paper "Which to Trust: Extracting Collective Wisdom Based on Opinion Quality Rank Learning".
- 2 December 2025
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in INFORMS Journal on Computing
- Vol. 38 (5) , 1673-1692
Which to Trust: Extracting Collective Wisdom Based on Opinion Quality Rank Learning
Aggregating diverse human opinions in the digital era is essential in harnessing collective wisdom to unravel intricate management dilemmas. The aggregation process has significant hurdles to overcome due to the heterogeneity in the quality of individual opinions. This study introduces a CrowdRank machine learning architecture to provide an innovative solution to the central problem of opinion aggregation by learning to rank individual opinions. CrowdRank operates through two phases. (1) It leverages a BNN, pretrained on historical data, to conduct pairwise opinion comparisons. This network, designed to capture meaningful interactions between opinion features, adheres to key axiomatic principles to ensure a principled evaluation of opinion quality. (2) CrowdRank employs expectation propagation to synthesize these microassessments into a coherent global ranking of individual opinions. We validated the efficacy of our approach through a large-scale empirical investigation using real-world financial analyst forecasts. The validation results demonstrated the superiority of our approach over existing methods in accurately predicting both pairwise and aggregate opinion rankings. Importantly, CrowdRank significantly improves the objectivity and precision of collective financial analyst forecasts. This study contributes a theoretically robust and practically validated innovation to opinion aggregation and charts a new path in the application of machine learning to enhance the synthesis of collective wisdom. History: Accepted by Ram Ramesh, Area Editor for Data Science and Machine Learning. Funding: This work was partly supported by the National Natural Science Foundation of China [Grants 92370204 and 71974031]. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2024.0835 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2024.0835 ). The complete IJOC Software and Data Repository is available at https://informsjoc.github.io/ .
- 2 December 2025
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in INFORMS Journal on Computing
- Vol. 38 (5) , 1673-1692
Which to Trust: Extracting Collective Wisdom Based on Opinion Quality Rank Learning
Aggregating diverse human opinions in the digital era is essential in harnessing collective wisdom to unravel intricate management dilemmas. The aggregation process has significant hurdles to overcome due to the heterogeneity in the quality of individual opinions. This study introduces a CrowdRank machine learning architecture to provide an innovative solution to the central problem of opinion aggregation by learning to rank individual opinions. CrowdRank operates through two phases. (1) It leverages a BNN, pretrained on historical data, to conduct pairwise opinion comparisons. This network, designed to capture meaningful interactions between opinion features, adheres to key axiomatic principles to ensure a principled evaluation of opinion quality. (2) CrowdRank employs expectation propagation to synthesize these microassessments into a coherent global ranking of individual opinions. We validated the efficacy of our approach through a large-scale empirical investigation using real-world financial analyst forecasts. The validation results demonstrated the superiority of our approach over existing methods in accurately predicting both pairwise and aggregate opinion rankings. Importantly, CrowdRank significantly improves the objectivity and precision of collective financial analyst forecasts. This study contributes a theoretically robust and practically validated innovation to opinion aggregation and charts a new path in the application of machine learning to enhance the synthesis of collective wisdom. History: Accepted by Ram Ramesh, Area Editor for Data Science and Machine Learning. Funding: This work was partly supported by the National Natural Science Foundation of China [Grants 92370204 and 71974031]. Supplemental Material: The software that supports the findings of this study is available within the paper and its Supplemental Information ( https://pubsonline.informs.org/doi/suppl/10.1287/ijoc.2024.0835 ) as well as from the IJOC GitHub software repository ( https://github.com/INFORMSJoC/2024.0835 ). The complete IJOC Software and Data Repository is available at https://informsjoc.github.io/ .
- 28 August 2024
- Published by Elsevier in European Journal of Operational Research
- Vol. 320 (3) , 720-738
Multiple financial analyst opinions aggregation based on uncertainty-aware quality evaluation
- 28 August 2024
- Published by Elsevier in European Journal of Operational Research
- Vol. 320 (3) , 720-738
Multiple financial analyst opinions aggregation based on uncertainty-aware quality evaluation