Sujian Li

Sujian Li is an active researcher in computational linguistics and natural language processing, with recent contributions to advanced large language model applications. Their work spans multiple critical areas including hierarchical memory frameworks for Wikipedia generation, self-refining entity grounding systems, and long-context embedding model extensions.

• Key Research Areas: Continual learning in NLP, multimodal reasoning, cross-lingual knowledge transfer, and factual consistency evaluation.
• Notable Methods: MOG framework for structured generation, ISR self-refinement scheme, LongAttn token-level analysis, and IPR step-level process refinement.

Article Trends show a focus on improving LLM robustness through adversarial training, enhancing coherence via discourse-level graph modeling, and developing benchmarks like WIKIGENBENCH for real-world evaluation. Their research also addresses knowledge integration in biomedical multilingual models (KBioXLM) and mathematical parsing via tree-structured decoding.

Collaborations include leading researchers like Yifan Song, Dawei Zhu, and Wenhao Wu across institutions and projects.