Caijun Xu

I am currently a Ph.D. student at Fudan University & SII, where I am a member of Alex Research, working under the supervision of Prof. Yixin Cao. My primary research interests lie in data synthesis and reasoning for large language models (LLMs), as well as the application of reinforcement learning (RL) in these areas. Prior to my doctoral studies, I obtained a B.Eng. from Beihang University (BUAA), where I worked on recommender systems, with a focus on Graph Neural Networks (GNNs) and Knowledge Graphs for enhancing recommendation algorithms.

My research involves the development of novel techniques for data synthesis, which aim to improve the robustness and generalization of models. This includes creating synthetic data pipelines and leveraging data augmentation methods to support large-scale training. In the realm of reasoning, I am exploring ways to enhance LLMs by integrating multi-step reasoning capabilities using RL, focusing on trajectory generation, feedback mechanisms, and evaluative verification processes. These efforts aim to advance the state of LLMs in practical applications that require complex reasoning tasks.

Additionally, during my undergraduate studies at BUAA, I contributed to the development of recommender systems, specifically by applying GNNs and knowledge graph-based approaches to optimize recommendation strategies. This research focused on improving the accuracy and relevance of recommendations in environments with sparse data.

For more information, you can access my Google Scholar, GitHub. Feel free to contact me via email at xxcaijun@gmail.com.