iCASE PhD Studentship with QinetiQ¶
This studentship focuses on assurance methods for large language models. The research explores how auxiliary knowledge can reveal hallucinations, falsehoods, contradictions, and gaps in model outputs, while also examining how LLM inference, training, and architecture can be improved for stronger correctness.
Project Details¶
| Item | Detail |
|---|---|
| Type | Studentship |
| Status | Live |
| Funder | Engineering and Physical Sciences Research Council |
| Value | £42,000 |
| Dates | 1 January 2024 - 31 December 2027 |
| Partner | Naimuri, a QinetiQ company |
| Koorosh's role | Co-Investigator / supervisor |
Project Focus¶
- LLM assurance through auxiliary and structured knowledge.
- Detecting hallucinations, contradictions, falsehoods, and missing knowledge.
- Improving model reliability across inference, training, and architectural choices.
- Building evidence for dependable use of LLMs in sensitive domains.