Yihong Chen
Dr Yihong Chen
Department of Computer Science,
Parks Road 14/15,
Room 40.02
Directions Postal Address
Interests
I seek to understand intelligence and its computational reproduction, informed by human cognition. A central aspect of this pursuit is the acquisition of knowledge: how it is formed, used in daily reasoning, and extrapolated to solve tasks in the wild. The ultimate goal would be to build a knowledge engine. Within this ultimate goal, I study different paradigms, including structured approaches such as knowledge graphs and graph neural networks, as well as unstructured approaches such as language models. Methodology-wise, I am particularly drawn to learning processes that shape knowledge over long time horizons, with meta-learning serving as a means of abstraction and continual learning as a mechanism for the gradual integration of knowledge. Through these lenses, I aim to understand how AI systems can learn to efficiently abstract, represent, and use concepts or symbols. To ground these ideas, my research interests extend to concrete applications, including recommender systems, search engines, chatbots, and general AI assistants.
Biography
I am a Postdoctoral Researcher in the Department of Computer Science at the University of Oxford, affiliated with the Oxford Applied and Theoretical Machine Learning (OATML) group, where I am supervised by Yarin Gal. I completed my Ph.D. in Computer Science at University College London through the FAIR–UCL PhD Programme, jointly hosted with Meta AI, under the supervision of Sebastian Riedel and Pontus Stenetorp. During my doctoral training, I was based at both UCL and Meta AI FAIR, working across academic and industrial research environments. I hold both a master’s and a bachelor’s degree in Electronic Engineering from Tsinghua University, and have undertaken research roles at Microsoft Research Asia (MSRA) and Huawei Noah’s Ark Lab. My research has been supported by the ELLIS Ph.D. Programme, the Meta AI FAIR Ph.D. Scholarship, the DeepMind Studentship at the UCL Centre for Doctoral Training in Foundational Artificial Intelligence (FAI CDT), the Google Women Techmakers Scholarship etc. I contribute to open-source research software (GitHub), and some of my work has reached a broader audience through public scientific writing, including an interview with Quanta Magazine on selective forgetting in AI. In general, I choose my research direction based on curiosity and seek to create technology where scientific rationality meets the human heart.
Selected Publications
-
Structure and Destructure: Dual Forces in the Making of Knowledge Engines
Yihong Chen
PhD Thesis University College London. 2025.
Details about Structure and Destructure: Dual Forces in the Making of Knowledge Engines | BibTeX data for Structure and Destructure: Dual Forces in the Making of Knowledge Engines | Link to Structure and Destructure: Dual Forces in the Making of Knowledge Engines
-
Improving Language Plasticity via Pretraining with Active Forgetting
Yihong Chen‚ Kelly Marchisio‚ Roberta Raileanu‚ David Ifeoluwa Adelani‚ Pontus Stenetorp‚ Sebastian Riedel and Mikel Artetxe
In NeurIPS 2023‚ Thirty−seventh Conference on Neural Information Processing Systems. 2023.
Details about Improving Language Plasticity via Pretraining with Active Forgetting | BibTeX data for Improving Language Plasticity via Pretraining with Active Forgetting
-
ReFactor GNNs: Revisiting Factorisation−based Models from a Message−Passing Perspective
Yihong Chen‚ Pushkar Mishra‚ Luca Franceschi‚ Pasquale Minervini‚ Pontus Stenetorp and Sebastian Riedel
In NeurIPS 2022‚ Thirty−sixth Conference on Neural Information Processing Systems. 2022.
Details about ReFactor GNNs: Revisiting Factorisation−based Models from a Message−Passing Perspective | BibTeX data for ReFactor GNNs: Revisiting Factorisation−based Models from a Message−Passing Perspective