Publications

2026

ICML'26FactGuard: Agentic Video Misinformation Detection via Reinforcement Learning

Zehao Li, Hongwei Yu, Hao Jiang, Qiang Sheng, Yilong Xu, Baolong Bi, Yang Li, Zhenlong Yuan, Yujun Cai, Zhaoqi Wang

Proceedings of the 43rd International Conference on Machine Learning (Acceptance Rate: 6352/23918=26.6%)

We propose an agentic misinformation video detection framework that can reason the veracity iteratively with self-refinement.

Preprint

ACL'26Logical Consistency as a Bridge: Improving LLM Hallucination Detection via Label Constraint Modeling between Responses and Self-Judgments

Hao Mi, Qiang Sheng, Shaofei Wang, Beizhe Hu, Yifan Sun, Zhengjia Wang, Hengqi Zeng, Yang Li, Danding Wang, Juan Cao

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Acceptance Rate: 19%)

We model the logical label constraint between LLM responses and self-judgments as a bridge to enhance hallucination detection.

Preprint

ACL'26Beyond the Final Actor: Modeling the Dual Roles of Creator and Editor for Fine-Grained LLM-Generated Text Detection

Yang Li, Qiang Sheng, Zhengjia Wang, Yehan Yang, Danding Wang, Juan Cao

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Acceptance Rate: 19%)

We disentangle the dual roles of creator and editor in LLM text generation, enabling fine-grained detection of AI-generated content at different revision modes.

IEEE TVCGInteractive Visual Assessment for Text-to-Image Generation Models

Xiaoyue Mi, Fan Tang, Juan Cao, Qiang Sheng, Ziyao Huang, Peng Li, Yang Liu, Tong-Yee Lee

IEEE Transactions on Visualization and Computer Graphics

We build an interactive visual assessment tool for exposing T2I models' vulnerability.

EACL'26Tailoring Rumor Debunking to You: Diversifying Chinese Rumor-Debunking Passages with an LLM-Driven Simulated Feedback-Enhanced Framework

Xinle Pang, Danding Wang, Qiang Sheng, Yifan Sun, Beizhe Hu, Juan Cao

Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics

We build an LLM-driven framework that generates rumor-debunking passages for various groups of people.

Paper

AAAI'26Reasoning About the Unsaid: Misinformation Detection with Omission-Aware Graph Inference

Zhengjia Wang, Danding Wang, Qiang Sheng, Jiaying Wu, Juan Cao

Proceedings of the 40th AAAI Conference on Artificial Intelligence (Acceptance Rate: 4167/23680=17.60%)

We consider the omitted information to better reason the creator's intent for misinformation detection.

2025

NeurIPS'25From Judgment to Interference: Early Stopping LLM Harmful Outputs via Streaming Content Monitoring

Yang Li, Qiang Sheng, Yehan Yang, Xueyao Zhang, Juan Cao

Proceedings of the 39th Annual Conference on Neural Information Processing Systems (Acceptance Rate: 5290/21575=24.52%)

We build a content moderator that can early stop LLMs' harmful outputs with low latency.

EMNLP'25The Staircase of Ethics: Probing LLM Value Priorities through Multi-Step Induction to Complex Moral Dilemmas

Ya Wu, Qiang Sheng, Danding Wang, Guang Yang, Yifan Sun, Zhengjia Wang, Yuyan Bu, Juan Cao

Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing (Acceptance Rate: 1811/8174=22.16%)

We build the MMD benchmark to probe the LLMs' value preferences in complex moral dilemmas.

EMNLP'25 FindingsForewarned is Forearmed: Pre-Synthesizing Jailbreak-like Instructions to Enhance LLM Safety Guardrail to Potential Attacks

Sheng Liu, Qiang Sheng, Danding Wang, Yang Li, Guang Yang, Juan Cao

Findings of the Association for Computational Linguistics: EMNLP 2025 (Acceptance Rate: 1417/8174=17.34%)

We enhance LLM resistance to jailbreaking by synthesizing instructions that may fall into the potentially risky space.

CIKM'25Bridging Thoughts and Words: Graph-Based Intent-Semantic Joint Learning for Fake News Detection

Zhengjia Wang, Qiang Sheng, Danding Wang, Beizhe Hu, Juan Cao

Proceedings of the 34th ACM International Conference on Information and Knowledge Management (Acceptance Rate: 443/1627=27.2%)

We propose to inject intent information with graph-based joint learning into fake news detection.

CIKM'25Enhancing Fake News Video Detection via LLM-Driven Creative Process Simulation

Yuyan Bu, Qiang Sheng, Juan Cao, Shaofei Wang, Peng Qi, Yuhui Shi, Beizhe Hu

Proceedings of the 34th ACM International Conference on Information and Knowledge Management (Acceptance Rate: 185/604=30.6%)

We propose AgentAug, which simulates the creative process of fake news videos using LLMs to mitigate the data scarcity issue.

IP&MExploring News Intent and Its Application: A Theory-Driven Approach

Zhengjia Wang, Danding Wang, Qiang Sheng, Juan Cao, Siyuan Ma, Haonan Cheng

Information Processing & Management

We, for the first time, conceptualize and computerize news intent modeling and showcase its application on fake news detection, popularity prediction, and propaganda detection.

ACL'25 FindingsEnhancing the Comprehensibility of Text Explanations via Unsupervised Concept Discovery

Yifan Sun, Danding Wang, Qiang Sheng, Juan Cao, Jintao Li

Findings of the Association for Computational Linguistics: ACL 2025

We exploit LLMs to automatically discover key concepts for text classification to enhance the comprehensibility of text explanations.

SIGIR'25LLM-Generated Fake News Induces Truth Decay in News Ecosystem: A Case Study on Neural News Recommendation

Beizhe Hu, Qiang Sheng, Juan Cao, Yang Li, Danding Wang

Proceedings of The 48th International ACM SIGIR Conference on Research and Development in Information Retrieval (Acceptance Rate: 239/1071=22.3%)

We reveal the truth-decay phenomenon where real news gradually loses its top-ranked advantage against fake news when LLM-generated news penetrates.

CAS BulletinChallenges on Public Security System in AI Era

Juan Cao, Qiang Sheng, Guojie Li

Bulletin of Chinese Academy of Sciences

We discussed new challenges and countermeasures in public security brought by AIGC techniques.

Paper

FCSExploiting User Comments for Early Detection of Fake News Prior to Users' Commenting

Qiong Nan, Qiang Sheng, Juan Cao, Yongchun Zhu, Danding Wang, Guang Yang, Jintao Li

Frontiers of Computer Science

We explored how to let a comment-based model effectively help a content-only one in fake news early detection.

2024

CIKM'24Let Silence Speak: Enhancing Fake News Detection with Generated Comments from Large Language Models

Qiong Nan, Qiang Sheng, Juan Cao, Beizhe Hu, Danding Wang, Jintao Li

Proceedings of the 33rd ACM International Conference on Information and Knowledge Management (Acceptance Rate: 347/1496=23.2%)

We prompt LLMs to role-play social media users to obtain generated comments for enhancing fake news detection.

MM'24FakingRecipe: Detecting Fake News on Short Video Platforms from the Perspective of Creative Process

Yuyan Bu, Qiang Sheng, Juan Cao, Peng Qi, Danding Wang, Jintao Li

Proceedings of the 32nd ACM International Conference on Multimedia (Acceptance Rate: 1149/4385=26.2%)

We detect fake news on short video platforms by modeling videos from the faking process perspective and constructed a FakeSV's sister dataset in English, namely FakeTT.

IJCAI'24Ten Words Only Still Help: Improving Black-Box AI-Generated Text Detection via Proxy-Guided Efficient Re-Sampling

Yuhui Shi, Qiang Sheng, Juan Cao, Hao Mi, Beizhe Hu, Danding Wang

Proceedings of the 33rd International Joint Conference on Artificial Intelligence (Acceptance Rate: 791/5651=14.0%)

To detect and attribute text generated by black-box LMs, we estimate their generation probabilities of representative words guided by a white-box proxy LM to obtain a strong feature.

IJCAI'24Vision-fused Attack: Advancing Aggressive and Stealthy Adversarial Text against Neural Machine Translation

Yanni Xue, Haojie Hao, Jiakai Wang, Qiang Sheng, Renshuai Tao, Yu Liang, Pu Feng, Xianglong Liu

Proceedings of the 33rd International Joint Conference on Artificial Intelligence (Acceptance Rate: 791/5651=14.0%)

We propose a vision-fused attack framework to acquire powerful adversarial text in neural machine translation.

AAAI'24Bad Actor, Good Advisor: Exploring the Role of Large Language Models in Fake News Detection

Beizhe Hu, Qiang Sheng, Juan Cao, Yuhui Shi, Yang Li, Danding Wang, Peng Qi

Proceedings of the 38th AAAI Conference on Artificial Intelligence

Large LMs generally underperform fine-tuned Small LMs for fake news detection, but they can be good advisors by providing rationales.

2023

CCCFThe Safety in the AIGC era: Techniques Make the World More Credible

Juan Cao, Qiang Sheng

Communications of the China Computer Federation

Our perspective paper on the safety issues brought by AI-generated content.

Paper

MM'23Combating Online Misinformation Videos: Characterization, Detection, and Future Directions

Yuyan Bu, Qiang Sheng, Juan Cao, Peng Qi, Danding Wang, Jintao Li

Proceedings of the 31st ACM International Conference on Multimedia (Acceptance Rate: 902/3669=24.6%)

We conduct the very first survey focusing on online misinformation video detection, to the best of our knowledge.

ACL'23Learn over Past, Evolve for Future: Forecasting Temporal Trends for Fake News Detection

Beizhe Hu, Qiang Sheng, Juan Cao, Yongchun Zhu, Danding Wang, Zhengjia Wang, Zhiwei Jin

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 5: Industry Track)

We propose to address the temporal shift issue in real-world fake news detection systems via forecasting topic-level trends and accordingly adjusting the detector update strategy.

2022

COLING'22Improving Fake News Detection of Influential Domain via Domain- and Instance-Level Transfer

Qiong Nan, Danding Wang, Yongchun Zhu, Qiang Sheng, Yuhui Shi, Juan Cao, Jintao Li

Proceedings of the 29th International Conference on Computational Linguistics (Acceptance Rate: 632/2253=28.1%)

We propose a domain- and instance-level transfer framework for detecting fake news of target domains.

IEEE TKDEMemory-Guided Multi-View Multi-Domain Fake News Detection

Yongchun Zhu, Qiang Sheng, Juan Cao, Qiong Nan, Kai Shu, Minghui Wu, Jindong Wang, Fuzhen Zhuang

IEEE Transactions on Knowledge and Data Engineering

We tackle the issue of domain shift and domain labeling incompleteness for simultaneously modeling multi-domain fake news via multi-view encoding and memory bank mechanism.

IP&MCharacterizing multi-domain false news and underlying user effects on Chinese Weibo

Qiang Sheng, Juan Cao, H Russell Bernard, Kai Shu, Jintao Li, Huan Liu

Information Processing & Management

Based on the largest false news dataset on Chinese Weibo (44k+), we characterize the spread of false news in nine domains and analyze the underlying user effects.

SIGIR'22Generalizing to the Future: Mitigating Entity Bias in Fake News Detection

Yongchun Zhu, Qiang Sheng, Juan Cao, Shuokai Li, Danding Wang, Fuzhen Zhuang

Proceedings of the 45th International ACM SIGIR Conference on Research and Development in Information Retrieval (Acceptance Rate: 165/667=24.7%)

We reveal the entity bias in fake news detection datasets and propose a causal framework to mitigate such bias for better generalization to future data.

ACL'22Zoom Out and Observe: News Environment Perception for Fake News Detection

Qiang Sheng, Juan Cao, Xueyao Zhang, Rundong Li, Danding Wang, Yongchun Zhu

Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) (Acceptance Rate: 701/3378=20.75%)

For the first time, we propose to perceive signals from the news environment for fake news detection.

AAAI'22DRAG: Dynamic Region-Aware GCN for Privacy-Leaking Image Detection

Guang Yang, Juan Cao, Qiang Sheng, Peng Qi, Xirong Li, Jintao Li

Proceedings of the 36th AAAI Conference on Artificial Intelligence (Acceptance Rate: 1349/9020=15.0%)

We propose a GCN-based method to find out crucial privacy-indicative regions for privacy-leaking image detection dynamically.

2021

MM'21Improving Fake News Detection by Using an Entity-enhanced Framework to Fuse Diverse Multimodal Clues

Peng Qi, Juan Cao, Xirong Li, Huan Liu, Qiang Sheng, Xiaoyue Mi, Qin He, Yongbiao Lv, Chenyang Guo, Yingchao Yu

Proceedings of the 29th ACM International Conference on Multimedia

We point out three valuable text-image correlations in multimodal fake news and propose an entity-enhanced fusion framework for cross-modal correlation modeling in fake news detection.

CIKM'21Integrating Pattern-and Fact-based Fake News Detection via Model Preference Learning

Qiang Sheng, Xueyao Zhang, Juan Cao, Lei Zhong

Proceedings of the 30th ACM International Conference on Information and Knowledge Management (Acceptance Rate: 271/1251=21.7%)

We propose a graph-based model preference learning framework to separately handle the pattern and fact indicators in fake news detection.

ACL'21Article Reranking by Memory-Enhanced Key Sentence Matching for Detecting Previously Fact-Checked Claims

Qiang Sheng, Juan Cao, Xueyao Zhang, Xirong Li, Lei Zhong

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers) (Acceptance Rate: 571/2327=24.5%)

We detect previously fact-checked claims by matching them against the key sentences in fact-checking articles.

J-CRADSemantics-Enhanced Multi-modal Fake News Detection

Peng Qi, Juan Cao, Qiang Sheng

Journal of Computer Research and Development

We focus on the role of explicit semantics for fake news detection via visual semantic extraction and knowledge-guided multi-modal semantic interaction.

Paper

WWW'21Mining dual emotion for fake news detection

Xueyao Zhang, Juan Cao, Xirong Li, Qiang Sheng, Lei Zhong, Kai Shu

Proceedings of the Web Conference 2021 (Acceptance Rate: 357/1736=20.6%)

We leverage both publisher emotion and social emotion for fake news detection.

2020

ACL'20Integrating Semantic and Structural Information with Graph Convolutional Network for Controversy Detection

Lei Zhong, Juan Cao, Qiang Sheng, Junbo Guo, Ziang Wang

Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics (Acceptance Rate: 571/2244=25.4%)

We propose a GCN-based method for post-level controversy detection. Also, we release the first Chinese dataset for this task.

CCCFFake News and Disinformation Detection on the Internet: Recent Advances and Future Prospects

Juan Cao, Qiang Sheng, Peng Qi

Communications of the China Computer Federation

We review the recent advances of detecting fake news and disinformation and propose potential directions.

Springer Book ChapterExploring the Role of Visual Content in Fake News Detection

Juan Cao, Peng Qi, Qiang Sheng, Tianyun Yang, Junbo Guo, Jintao Li

Disinformation, misinformation, and fake news in social media

A survey chapter that focuses on the visual information for fake news detection.

Preprint