Cheng Xu

Research

My research centers on developing robust methods for detecting and combating misinformation in the digital age. I work at the intersection of natural language processing, deep learning, and information verification to build systems that can understand and analyze complex information across different modalities. Additionally, I am also interested in the safety, fairness, and evaluation of LLMs, as well as their application in the financial sector.

• Fake News Detection: Developing fuzzy deep learning systems that integrate multimodal signals to identify misinformation with high accuracy. My work focuses on creating interpretable models that can explain their detection decisions, especially for the LLMs.
• LLM Safety, Fairness, and Evaluation: Investigating how training data and its internal reasoning mechanisms affects LLM safety, fairness and evaluation, particularly in fake news detection tasks.
• Financial LLMs: Applying LLMs to the financial sector, including integrating information systems such as news into the LLM decision-making process. Further exploring the potential of LLMs in financial market analysis, investment, etc.