Jianghan Liu

I am Jianghan Liu, a Master's candidate in Software Engineering at Southeast University(2023 intake), advised by Assoc. Prof. Wenjun Ke . I have received the B.E. degree from Southwest University.

My research focuses on optimizing topic modeling for LLMs and innovating knowledge acquisition paradigms. I am particularly interested in exploring advanced methods to enhance model performance and scalability in complex scenarios.

Google Scholar  /  Email

• 2025-06 One paper is accepted by IEEE Transactions on Knowledge and Data Engineering (TKDE).
• 2025-05 Two papers are accepted as ACL 2025 Oral (Top 2.6%), focusing on mechanisms of novel knowledge learning in large language models and topic modeling in the era of large language models.

• 2025.5 ~ now: LLM-based Generative Recommendation Model at Meituan
  • Generative Retrieval: Responsible for reproduction and optimization of generative recommendation system models based on the Transformer architecture, aiming to enhance the model's generalization capability across multiple scenarios and propose effective solutions for the cold start problem.
  • Sparse Feature Fusion: Explore and implement various ways of integrating features (such as position encoding, timestamps, user profiles, etc.) with product sequences, and evaluate their impact on recommendation effectiveness, promoting increased generalization capability of the model.
  • Post-Training of Generative Models: Develop GRPO reinforcement learning code suitable for the Encoder-Decoder architecture, using scores from reranking models as rewards, combined with format rewards and correct-answer rewards, to further enhance the model's capabilities.