Anthony Hughes

Anthony Hughes

PhD Student, Computer Science
ajhughes3 at sheffield.ac.uk

I am a Ph.D. student at the University of Sheffield. This is supported by a fully-funded UKRI scholarship. Currently on leave from PhD studies as a SPAR AI safety fellow. I'm working Andrew Draganov on backdoor detection in LLMs.

Research: My PhD research investigates privacy in deep learning. With a particular focus on how language models inadvertently leak information, and how we mitigate this.

Recent work: My EMNLP 2025 paper investigates information leakage in abstractive summarisation. My EACL 2026 paper, uses activation patching and circuit discovery to understand and mitigate PII leakage in language models.

Selected Publications in Privacy

PATCH: Mitigating PII Leakage in Language Models with Privacy-Aware Targeted Circuit PatcHing
A Hughes, V Duddu, N Asokan, N Aletras, N Ma
EACL 2026 - Findings
How Private are Language Models in Abstractive Summarisation?
A Hughes, N Aletras, N Ma
EMNLP 2025 - Main Proceedings

Selected Publications in Health/Medicine

Large Language Models to Improve Understanding of Radiology Reports
S Alabed et al. including A Hughes.
The Lancet Digital Health
Identifying and Aligning Medical Claims Made on Social Media with Medical Evidence
A Hughes, Xingyi Song
LREC-COLING 2024
Understanding Inflicted Injuries in Young Children: Toward an Ontology-based Approach
F Maikore et al. including A Hughes
EKAW 2024

Research Experience

Visiting Graduate Researcher
Cryptography, Security and Privacy Group (CrySP), University of Waterloo
July 2025 - September 2025

Collaborators/Mentors: Vasisht Duddu, N. Asokan. Research on mechanistic approaches to understanding PII leakage in language models.

Selected Awards

UKRI PhD Scholarship
University of Sheffield, 2023-2027
NIHR/BRC Sheffield Funding
£1,650, April 2025 - September 2025
Best Poster
Insigneo Showcase, July 2025