Reliable and Efficient Adaptation of Large Language Models
University of Sheffield – School of Computer Science
Supervisors: Dr Cass Zhixue Zhao and Prof Nikos Aletras
About the Project
This PhD project develops methods for the reliable, explainable, and efficient post-training adaptation of large language models, ensuring behavioural changes are effective, persistent, and safe.
Large language models (LLMs) are often adapted after training through fine-tuning, knowledge editing, and activation steering. While efficient, their effects are fragile. Interventions designed to improve one capability may unintentionally alter unrelated knowledge, reasoning, fairness, or safety.
This PhD project will develop methods for the reliable, explainable, and efficient post-training adaptation of LLMs. The central aim is to understand how interventions modify a model’s internal representations, ensuring behavioural changes are effective, persistent, and safe.
Research questions include:
- Can intended and unintended changes be detected from internal representations?
- Can we design lightweight monitoring methods that distinguish genuine internal changes from superficial compliance?
- How can adaptation preserve reasoning, fairness, and safety under limited compute?
Approaches may include representation analysis, mechanistic interpretability, model editing, efficient fine-tuning, and adversarial evaluation. Expected outcomes include new algorithms for reliable adaptation, evaluation frameworks, open-source tools, and publications at leading ML/NLP venues.
Supervisor Bio
Dr Cass Zhixue Zhao is a lecturer in Natural Language Processing at the University of Sheffield, focusing on trustworthy NLP, model compression, editing, and steering. High-level, my interests include AI safety and exploring dynamic interactions between humans and multi-agents. According to CSRankings, Dr Cass Zhixue Zhao is among the most productive UK AI researchers since 2024. This project offers an exciting opportunity to ensure LLM adaptation is safe and reliable.
About the School/Research Group
99% of our research is rated in the highest two categories in the REF 2021, meaning it is world-leading or internationally excellent. We are rated 8th nationally for research environment quality. The successful applicant will join the University of Sheffield’s Natural Language Processing Group, one of the UK’s largest and longest-established NLP research groups, benefiting from a collaborative environment and high-performance computing resources.
Candidate Requirements
Applicants should have, or expect to obtain, a Undergraduate degree (at least a UK 2:1 honours degree, or its international equivalent) in Computer Science, AI, mathematics, or a related discipline.
Essential/desirable experience:
- Strong foundation in ML or NLP
- Python proficiency and PyTorch experience
- Familiarity with language models, Transformers, or generative AI
- Ability to conduct independent research
- Interest in trustworthy, explainable, or efficient AI
Prior publications are welcome but not required. Applicants from adjacent fields with strong mathematical/ML experience are encouraged to apply.
How to Apply
To apply for a PhD studentship, applications must be made directly to the University of Sheffield using the Postgraduate Online Application Form. Make sure you name Dr. Cass Zhao and Prof. Nikos Aletras as your proposed supervisor(s). Information on what documents are required and a link to the application form can be found here - https://www.sheffield.ac.uk/postgraduate/phd/apply/applying. The form has comprehensive instructions for you to follow, and pop-up help is available.
Your research proposal should:
- be no longer than 4 A4 pages, include references
- outline your reasons for applying for this studentship
- explain how you would approach the research, including details of your skills and experience in the topic area
Applicants should also submit a CV, academic transcripts, a brief research statement, and the contact details of two referees.
Funding Notes
The PhD studentship will fund the standard UK tuition fee and a tax-free stipend at the standard UKRI rate (currently £21,805 for the 2026/27 academic year) for 3.5 years, plus a research training support grant to help fund research-related training, travel, and conference attendance. If you are an overseas student, you are eligible to apply but you must have the means to pay the difference between the UK and Overseas tuition fees by securing additional funding or self-funding (you need to provide details on this in your proposal). Further information on Home and Overseas fees can be found here - View Website
This is a Preview Listing…
You must sign in to see the full job description, and to apply.
Manage / Upgrade this job to a Full Job Listing.
Find Your Best Opportunity
Tell them AcademicJobs.com sent you!
