PhD Scholarship in Uncertainty Quantification of LLMs

Overview of the Role

By scaling up data, compute and model size, large language models (LLMs) have gained an impressive and ever growing array of capabilities. The next phase of development will be dominated by the use of LLMs in real world scenarios, and improving LLM reliability will be a crucial and exciting component of this next phase.

A PhD scholarship is available in a joint project with University of Oxford (Prof Yee Whye TEH), Nanyang Technological University (Prof Luke Ong) and National University of Singapore (Prof LEE Wee Sun) studying the reliability of LLMs through the lens of uncertainty quantification (UQ), Bayesian inference, conformal prediction, and world models.

Along with industrial and government partners, the postholder and project team will use these methods to address current issues with LLMs, such as reward hacking, active learning and testing of LLMs.

Reporting to Prof Luke Ong, the postholder will be a member of his research group in the College of Computing and Data Science, NTU.

Selection Criteria

Applicants should be close to completion or hold a masters degree (or equivalent) in computer science or mathematics or statistics. The ideal candidate for the PhD scholarship should have prior research experience in statistical ML, Deep Learning and LLMs.

Applications and Enquiries

To apply, email a copy of your application to the PI, Luke Ong (Email: luke.ong @ ntu . edu . sg), with the subject: PhD Scholarship in Uncertainty Quantification of LLMs, including the following:

• your academic transcripts and CV
• representative papers or reports (if available)
• contacts of three references.

One PhD scholarship is available. This call was posted in January 2025. It will remain open until the position is filled. Only shortlisted candidates will be notified.