Summer Sale

Up to 30% off — ends 2 Aug

ONLY00h00m00s

Publisher
source

University of Cambridge

Closing soon

Top university

PhD Studentship in Monitoring and Increasing LLM Safety University of Cambridge in United Kingdom

Degree Level

PhD

Field of study

Computer Science

Funding

Full funding available

Deadline

Jul 30, 2026

Country flag

Country

United Kingdom

University

University of Cambridge

Social connections

How do I apply for this?

Sign in for free to reveal details, requirements, and source links.

Meet Kite AI

More information can be found here

Keywords

Computer Science
Psychology
Cognitive Science
Electrical Engineering
Information Technology
Mathematics
Interpretability
Statistics
Large Language Models
Machine learning

Suggested scholarships

About this position

[Fully funded studentship covering fees and maintenance for a home or overseas candidate.]

PhD Studentship in Monitoring and Increasing LLM Safety at the University of Cambridge.

This fully funded studentship supports research on making large language models safer, with a focus on evaluating model behaviour, monitoring outputs at inference time, and reducing risks from LLM deployment. The project is grounded in AI safety and explores approaches such as white-box mechanistic interpretability and black-box behavioural research.

Early work in the first 1.5 years will concentrate on increasing chain-of-thought faithfulness and mitigating encoded reasoning. Two possible research directions are outlined: one investigates whether CoT is used in a straightforward way by applying perturbation methods such as paraphrasing intermediate outputs and then examining performance changes; the other aims to train for transparency by using a human or AI predictor to assess whether model outputs can be predicted from the CoT, with predictor accuracy used as a reward signal during training.

The award is fully funded and covers fees and maintenance for either a home or overseas candidate. Funding is provided by Coefficient Giving, and the project is aligned with their focus on encoded reasoning in CoT and inter-model communication. After the initial scoped projects are completed, the student, supervisor, and funder will decide on the next research direction.

Applicants should have, or expect to obtain by the start date, at least a first degree in Engineering or a related subject. Experience in software development projects or research on LLMs is desirable. Candidates must also secure a place on the PhD programme through separate Graduate Admissions application to the University of Cambridge. The funding application is submitted via a Google Form with a two-page CV and research proposal, and applications are reviewed on a rolling basis. The postgraduate admissions deadlines noted in the advert are 14 May for October start and 30 July for January start, though earlier application is recommended.

This opportunity is a strong fit for students interested in machine learning safety, interpretability, reasoning transparency, and empirical evaluation of advanced AI systems.

Funding details

Full funding including tuition fees and living expenses is available for this position. The scholarship covers all educational costs and provides a monthly stipend.

How to apply

Please submit your application including a cover letter, CV, academic transcripts, and contact information for two references. Applications should be sent via the online portal before the deadline.

Ask ApplyKite AI

Start chatting
Can you summarize this position?
What qualifications are required for this position?
How should I prepare my application?