PhD Studentship, Monitoring and Increasing LLM Safety

The Effective Altruism

Opportunities Board

Work on the world's most pressing problems. Browse jobs, fellowships, internships, courses, and more at high-impact organisations.

Add opportunity

Get notified

Add opportunity

Get notified about new roles

Give us feedback →

PhD Studentship, Monitoring and Increasing LLM Safety

University of CambridgeCambridge, United Kingdom

Cambridge, United Kingdom

2 weeks ago

AI safety & policy

Funding

Fellowship

Deadline

2026-07-30

Routes to impact

Direct high impact on an important cause

Skill-building & building career capital

Testing your fit for a certain career path

Description

A fully-funded PhD studentship at the University of Cambridge (Department of Engineering), open to home and overseas candidates, focused on advancing the safety of large language models (LLMs) through interpretability and behavioural research.

Research focus - investigate and improve chain-of-thought (CoT) faithfulness and mitigate encoded reasoning using white-box mechanistic interpretability and black-box behavioural methods
Funding - fully funded by Coefficient Giving, covering both tuition fees and maintenance for the duration of the studentship
Projects - choose between two scoped research tracks: testing CoT transparency via perturbation methods, or training for reasoning transparency using a human predictor model
Eligibility - requires at least a first degree in Engineering or a related field; experience in software development or LLM research is desirable

Apply via the University's Graduate Admissions portal (closing dates: 14 May for October start, 30 July for January start); submit your CV and research proposal separately through the Coefficient Giving application form.

This text was generated by AI. If you notice any inconsistencies, please let us know using this form

View opportunity

Related opportunities