Jingjing Wang

Jingjing Wang

王晶晶

Associate Professor, Natural Language Processing Lab

School of Computer Science and Technology, Soochow University
djingwang@suda.edu.cn
No.1, Shizi Street, Suzhou, China, 215006
01

About

I am an Associate Professor at School of Computer Science and Technology, Soochow University. I am also a Senior Technical Consultant (Part-time) at Microsoft (Asia), China.

My research interests focus on Multimodal Computing (especially for Multimodal Information Extraction, Visual-Language Understanding and Generation, Embodied Intelligence), Affective Computing, Large Language Models and AI for Medical Diagnosis.

I received my Ph.D. degree from Soochow University in 2019, advised by Prof. Guodong Zhou and Prof. Shoushan Li. I also collaborate with Prof. Min Zhang for advancing NLP and AI technology.

王晶晶，苏州大学副教授，微软访问学者，苏州大学自然语言处理（NLP）实验室博士，兼任微软（亚洲）工程院高级技术顾问， 主要致力于人工智能（AI）领域中Multimodal Computing (especially for Multimodal Information Extraction, Visual-Language Understanding and Generation, Embodied Intelligence), Affective Computing, Large Language Models and AI for Medical Diagnosis等方向的研究。 截止目前，已在CCF-A类顶会和顶刊，例如ACL、AAAI、WWW、MM、IJCAI、SCIS等发表AI与NLP相关论文数十篇，并主持与参与国家项目多项，拥有多项授权专利。 此外担任AI、NLP领域国际顶级会议ACL、AAAI、WWW、MM、IJCAI等的Area Chair、PC，CCCF期刊编委以及TASLP、TAFFC、SCIS、中国科学、软件学报等国内外重要学术期刊审稿人。 目前合作的企业包括：Microsoft、阿里达摩院、蚂蚁金服等，也乐于推荐本组的学生到上述企业交流、实习与工作。人生寄语：“知者行之始，行者知之成”。

Welcome highly-motivated students to join my team. Perspective candidates are welcome to email me with your CV or research interests for detailed consultation. Regarding recommendation letters, please be advised that I would like to provide substantive evaluations for candidates with whom I have already had a meaningful collaboration. This would allow me to objectively assess your research competencies, scholarly contributions, and professional development through sustained engagement.

02

Updates & News

03

Selected Research

Multi-Agent System

Multi-Agent System for Structured Information Extraction and Generation

This research focuses on advancing Text-to-SQL parsing by addressing the critical challenge of semantic accuracy in LLM-generated SQL queries. While fine-tuned large language models excel at producing syntactically valid SQL, they often struggle with semantic consistency, leading to unreliable results in real-world applications.

To tackle this issue, we introduce SQLFixAgent (Cen et al., AAAI'2025), a novel multi-agent collaborative framework designed to detect and repair erroneous SQL queries. SQLFixAgent integrates three specialized agents:

1) SQLReviewer: Identifies semantic mismatches using rubber duck debugging principles.
2) QueryCrafter: Generates diverse candidate SQL repairs by perturbing user queries.
3) SQLRefiner: Selects the optimal repair through retrieval-augmented reflection and failure memory.

This framework achieves a 3% improvement in execution accuracy on the challenging BIRD benchmark while maintaining high token efficiency, making it practical for deployment. Beyond this, we investigate robust Text-to-SQL parsing across diverse scenarios, including domain knowledge integration (Spider-DK) and synonym robustness (Spider-Syn). Our work also explores the synergy between fine-tuned and foundation LLMs, demonstrating how agent collaboration can compensate for individual model limitations.

In addition, we propose Table-Critic (Yu et al., ACL'2025), a novel multi-agent framework designed to enhance structured table reasoning through collaborative error detection and refinement. While large language models (LLMs) excel in many reasoning tasks, they often struggle with maintaining consistency in multi-step table-based reasoning, leading to cascading errors. Table-Critic demonstrates how structured collaboration among LLM-based agents can overcome inherent limitations in complex reasoning tasks, offering a scalable and interpretable approach for real-world applications in data analysis and decision support.

SQLFixAgent PDF SQLFixAgent Code Table-Critic PDF Table-Critic Code
Video Understanding

Multimodal Foundation Model for Video Understanding and Grounding

The goal is to establish a unified framework for video anomaly detection, advancing precision in identifying and localizing abnormal events across dynamic scenes while enabling interpretable analysis of complex visual patterns.

Starting from real-world applications in surveillance and social media analysis, we introduce Hawkeye (Zhao et al., ACM MM'24), the first scene-enhanced video-language model designed for anomaly detection. Hawkeye integrates multimodal context (visual-textual-temporal cues) to recognize subtle anomalies and pinpoint their temporal boundaries in untrimmed videos. This work lays a critical foundation for event typing and spatiotemporal localization in short video understanding.

Building on this, we investigate low-resource scenarios where annotated anomaly data is scarce. Our Continuous Attention Modeling method (Zhang et al., JOS'23) enhances adaptability by capturing long-range dependencies in sparse anomaly signals. Further, we extend Hawkeye with self-supervised learning to uncover latent patterns across unlabeled videos, improving generalization to unseen anomaly types. To scale solutions, we construct a benchmark suite combining large-scale anomaly annotations and instruction-tuned datasets. This addresses the challenge of diverse event types (e.g., accidents, unusual behaviors) and supports downstream tasks like explainable reasoning.

Zhang et al., JOS'23 PDF Zhao et al., ACM MM'24 ACM Hawkeye Code Hawkeye PDF
04

Publications

2025
2024
2023
2022
2021
2020
2019
2018
2017
2015
05

Awards & Honors

Outstanding Expert and Supervisor of Microsoft (2021)
Outstanding PhD of Soochow University (2019)
Suzhou Industrial Park Scholarship (2018)
National Scholarship for Ph.D. (2017)
Ph.D. Scholarship of Soochow University (2017)
National Scholarship for Master (2016)
Outstanding Graduate Student of Soochow University (2016)
Suzhou Industrial Park Scholarship (2015)
06

Academic Services

Technical Program Committee

  • ACL: Annual Meeting of the Association for Computational Linguistics, Area Chair
  • EMNLP: Conference on Empirical Methods in Natural Language Processing, Area Chair
  • AAAI: Association for the Advancement of Artificial Intelligence, PC
  • IJCAI: International Joint Conference on Artificial Intelligence, PC
  • WWW, ACM MM: Reviewer

Journal Reviewer

  • IEEE/ACM TASLP
  • ACM TALLIP
  • Science China Information Sciences
  • Science China
  • Acta Automatica Sinica
  • Journal of Chinese Information Processing

Academic Presentations and Exchanges

  • 2016-2021: Academic reports and exchanges at top conferences including ACL, AAAI, IJCAI
  • 2019: Academic report and exchange at Zhejiang Tailong Commercial Bank, Suzhou Industrial Park Headquarters
  • 2019: Invited talk at Ecovacs, Suzhou
  • 2022: Academic report and exchange at Alibaba Ant Financial
  • 2023: Academic report and exchange at the establishment of NLPAI-SCHOOL, Microsoft Asia Engineering Institute, Suzhou
07

Research Grants

As Principle Investigator

  • Key Technologies for Multimodal Implicit Sentiment Understanding and Editing Generation in Harmful Short-Video Content Governance
    No. 62576234 · 500K RMB · 2026.01–2029.12
    NSFC General Program
  • Key Technology Research on Attribute-level Sentiment Analysis for Conversational Texts
    No. 62006166 · 240K RMB · 2021.01–2023.12
    NSFC Young Scientist Fund Project
  • Research on Chinese Single-document Automatic Summarization Based on Discourse Structure Analysis
    No. 61976146 · 560K RMB · 2020.01–2023.12
    NSFC General Program
  • Resource Construction and Key Technology Research on Sentiment Information Extraction from Question-answer Texts
    No. 2019M661930 · 80K RMB · 2020.01–2022.12
    China Postdoctoral Science Foundation (CPSF)

As Co-investigator

  • Scene-based Knowledge Graph for Language Understanding and Generation
    Sub-project No. 2020AAA0108604 · 6,650K RMB · 2020.11–2023.10
    National Key Research and Development Program