RG

PhD candidate · Queen Mary University of London

Roksana
Goworek

I pry open language models to see how they hold meaning. How it shifts across time, aligns across languages, and transfers across domains.

I'm a final-year PhD student at Queen Mary University of London, supervised by Haim Dubossarsky. I work on how language models represent meaning — and on what those representations capture, distort, and miss.

My research runs across mechanistic interpretability, representation modelling, lexical semantic change detection, cross-lingual transfer for low-resource languages, and information retrieval. The thread tying them together is a single question: how is meaning encoded inside a model, and how does it move — across senses, across time, and across languages?

Publications

Full list on Google Scholar →

Updates

  1. Jun–Nov 2026 Research internship at the Alan Turing Institute — mechanistic interpretability
  2. Jun–Jul 2026 BlueDot technical AI-safety course
  3. Jun 2026 Entered BlueDot's technical AI-safety puzzle
  4. Mar 2026 Published my LSCD metrics paper at LChange 2026
  5. Jan 2026 Attended the 2nd UniDive training school, Yerevan
  6. Nov 2025 Presented at EMNLP 2025, Suzhou — a main-conference paper and a co-authored paper
  7. Jun–Nov 2025 Research internship at the Alan Turing Institute — cross-lingual information retrieval
  8. Sep 2025 4th Change is Key! consortium, Uppsala & Gothenburg
  9. Aug 2025 SenWiCh presented at the SIGTYP workshop, ACL 2025
  10. Aug 2025 Work presented at Societas Linguistica Europaea (SLE 2025)
  11. Sep 2023 – May 2025 Freelance AI software developer, Copylab
  12. Sep 2024 3rd Change is Key! consortium, London
  13. Jul 2024 Designed & led a session at the ESSLLI 2024 lexical semantic change workshop
  14. Mar 2024 Presented my first paper at EACL 2024
  15. Oct 2023 2nd Change is Key! consortium
  16. Sep 2023 Started my PhD at Queen Mary University of London
  17. Jun–Sep 2023 AI software developer internship, Copylab
  18. Jul 2023 Graduated Edinburgh — First-Class BSc in Artificial Intelligence

Demos & Tools

tool

SemLens

An interactive tool for exploring lexical semantic change — how a word's usage and meaning drift over time.

notebook

Contrastive Learning

A walkthrough notebook on contrastive representation learning.

Elsewhere