Zhejiang University NLP

About

Zhejiang University's Natural Language Processing (NLP) group, commonly recognized as ZJUKG (Zhejiang University Knowledge Graph), plays a pivotal role in advancing research in generative AI and Large Language Models (LLMs). Although the group itself wasn't founded in 1897, its host, Zhejiang University, has a long-standing tradition of excellence in research and academia. ZJUKG leverages this legacy to delve into cutting-edge NLP research, particularly focusing on areas such as Knowledge Graphs (KGs) and LLMs. Their goal is to not only develop theoretical insights but also contribute practical solutions, especially in AI for scientific discovery. Central to ZJUKG's research agenda are several innovative projects and open-source initiatives. Notably, KnowLM, an open-source framework, stands out for its contributions to large-scale knowledgeable language models. This framework exemplifies their dedication to advancing LLMs by integrating comprehensive knowledge sources. Another significant project is DeepKE, which serves as an open tool for knowledge extraction, highlighting the group’s emphasis on making AI applications more accessible and effective. Additionally, ZJUKG has collaborated with Alibaba to create OpenBG, an open business KG, showcasing the practical applications of their research in real-world scenarios. One of the group’s primary research focuses is on improving the functionality and efficiency of LLMs through various methodologies. This includes automatic KG construction, neural reasoning, and innovative model editing techniques. Their work on EasyEdit, presented at ACL 2024, is a noteworthy example of how they enhance LLMs by allowing easier knowledge integration. Moreover, their publication on world knowledge models (WKMs) for agent planning in NeurIPS 2024 and studies on knowledge unlearning in EMNLP 2024 Findings underscore their commitment to addressing challenges like hallucination mitigation and enhancing reasoning abilities in LLMs. Beyond these technical achievements, ZJUKG is deeply involved in global collaboration, partnering with major industry players like Alibaba and esteemed academic institutions such as Oxford and Edinburgh Universities. This international collaboration not only enriches their research but also solidifies their status as leaders in NLP and AI research. The group's accolades, including best paper awards from CCKS 2021 and IJCKG 2021, further highlight their substantial contributions to the fields of knowledge graphs and language technologies. While more detailed insights into their specific work in generative AI might require additional sources, their emphasis on LLMs, knowledge integration, and model advancements positions them as significant contributors to the rapidly growing domain of generative AI.