EPSRC iCASE funded PhD studentship on Assuring Large Language Models Through Knowledge Graph Infusion

Please feel free to reshare.

About this PhD funded project
-------------------------------
PhD application deadline: 31st March 2024.

Recently, Large Language Models (LLMs) have seen huge advancement. QinetiQ (
https://www.qinetiq.com/en/) is developing an AI Assurance Framework, to
allow such tools to be responsibly utilised by UK Defence and other
government agencies. Assuring LLMs is a complex and rapidly evolving field,
where trustworthiness and factuality are less well understood than more
mature AI technologies.

This project aims to assure LLM products and improve their performance
using Neuro-Symbolic AI. Research in Neuro-Symbolic AI focuses on combining
neural network-based technologies (such as Deep Learning and LLMs) with
symbolic systems like Knowledge Graphs. Over five stages, we will explore
how to use knowledge graphs to detect hallucinations, falsehoods,
contradictions, and knowledge gaps in LLM outputs. Additionally, the
project will enhance LLM inference, training, and architecture to bolster
model quality. Through this, we aim to establish robust mechanisms for LLM
assurance and improvement.

This PhD in Computer Science project sits within the Responsible Artificial
Intelligence Group (https://www.responsibleaihull.com) at the University of
Hull, UK. The group is renowned for its work on Neuro-Symbolic AI. It
consists of 7 staff members and over 10 PhD students working in areas
relevant to this project. As a PhD student, you will join a growing
research community in this field. This is a prestigious EPSRC (Engineering
and Physical Sciences Research Council) iCASE studentship co-funded
by Naimuri (A QinetiQ subsidiary; https://naimuri.com).

As a PhD student, you will be expected to actively participate within the
programme of study, showing good time management and organisation. You are
expected to work upon and further develop the research questions posed
within the research proposal, completing tasks required to gain a PhD such
as attending meetings, regularly reviewing literature, completing pertinent
studies, disseminating their work at appropriate forums and writing a
thesis on the topic.

Supervisors (Academic):
<https://www.linkedin.com/in/ACoAAACSwPwBfmZC9VAr9PY3x3QgdzBnOTeBO1Y>Professor
Dhaval Thakker
<https://www.linkedin.com/in/professor-dhaval-thakker-9382793/>,
<https://www.linkedin.com/in/ACoAAAdMDjEBQC576WOReF6ZrKUxGYhq4tZ0aTg>Koorosh
Aslansefat <https://www.linkedin.com/in/koorosh-aslansefat/>
<https://www.linkedin.com/in/ACoAADM7UOYBqj5NSve7a6VUjAjtThqIUF5skhA>Kevin
Pimbblet <https://www.linkedin.com/in/kevin-pimbblet-606a2a1bb/>
<https://www.linkedin.com/in/ACoAAALhp9QBrh5j_Uofne8hREKti8KymOfu91c>Rajiv
Ranjan <https://www.linkedin.com/in/rajiv-ranjan-33023414/>
Industry (QinetiQ <https://www.linkedin.com/company/qinetiq_2/>/Naimuri
<https://www.linkedin.com/company/naimuri/>):
<https://www.linkedin.com/in/ACoAABlVhuEBbTaIZUu7_iishokNkVCEH7Cib0Y>Rowan
Morris <https://www.linkedin.com/in/rowan-morris-741415ba/>,
<https://www.linkedin.com/in/ACoAADjbA1oBuN63XP5U5wdQ3nsAUrq3uzcgFAo>James
Ramsden <https://www.linkedin.com/in/james-ramsden-422550227/>

Apply through:
https://www.hull.ac.uk/study/postgraduate/research/phd/funded/epsrc-icase-studentship-on-assuring-large-language-models-through-knowledge-graph-infusion

Thank you

--------------------




*Professor Dhaval ThakkerProfessor of Artificial Intelligence (AI) and
Internet of Things (IoT)Director of Research | Turing Academic Lead Lead of
Responsible Artificial Intelligence (RAI) Group*

*UoH PI on the EPSRC-funded National Edge AI Hub: *https://edgeaihub.co.uk


*School of Computer Science, Faculty of Science and Engineering University
of Hull*Hull, HU6 7RX, UK
www.hull.ac.uk
D.Thakker@hull.ac.uk

Follow me on Twitter: prof_thakker <https://twitter.com/prof_thakker>
Learn more about our RAI Group: Responsible AI Hull
<https://www.responsibleaihull.com/>

Learn more about our Smart Cities research: Smart Bradford
<http://smartbradford.uk/>