Robust and Efficient Rumor Detection on Social Media

The School of EECS is hosting the following HDR Progress Review 1 Confirmation Seminar:

Robust and Efficient Rumor Detection on Social Media

Speaker: Wei Jiang

Host/Chair: Dr Rocky Chen

Abstract

The widespread adoption of social media has significantly improved information accessibility while simultaneously facilitating the rapid spread of misinformation, particularly malicious rumors. These rumors propagate through complex user interactions, posing challenges for timely and accurate detection. This research addresses three core challenges in rumor detection: (1) the difficulty of ensuring robust performance across varying propagation structures due to inconsistent data quality; (2) the inefficiency of current early detection methods that rely on fixed time intervals, ignoring dynamic user behavior; and (3) the under-utilization of task-specific semantic knowledge, such as stance, sarcasm, and hate speech, in existing large language model (LLM)-based detectors. To address these challenges, we propose a multi-faceted approach: we incorporate epidemiological modeling to improve detection robustness and generalize it to a unified cross-task framework; we design an adaptive LLM-agent capable of dynamically determining the optimal time for early intervention; and we enable effective knowledge reuse by merging task-specific LLMs to enrich semantic understanding without the need for redundant training. These contributions aim to build more robust, efficient, and semantically aware rumor detection systems.

Bio

Wei Jiang is a PhD student in the School of Electrical Engineering and Computer Science at the University of Queensland. He earned his Master of Data Science degrees from UQ. His research centres on social media data mining, under the supervision of Dr Rocky Chen and Professor Hongzhi Yin.