About

I am a PhD student in Natural Language Processing at the University of Edinburgh (Sept 2025 - present) working under the supervision of Björn Ross, Emily Allaway, and Ivan Titov. I research how language models encode and represent generalizations in language (e.g., "PhD students are curious"), with application areas including detecting stereotypes (generalizations about groups of people) and exaggerated claims in scientific communication.

Before my PhD, I worked as a software engineer with Amazon's Alexa Speech team in Cambridge, UK (2022-2025). I have a Masters in Speech and Language Processing from the University of Edinburgh (2021-2022) and a Bachelors in Mathematics and Computer Science (minor in Linguistics) from the University of Maryland, College Park (2017-2021).

Selected publications

Full list on Google Scholar.

• The Missing Minimal Pair: Stereotype Evaluation in LLMs
  Nataliya Stepanova, Ivan Titov, Emily Allaway, Björn Ross
  Under review, 2026
  Paper Code

  BibTeX

  @misc{stepanova2026missingminimalpairstereotype,
  title = {The Missing Minimal Pair: Stereotype Evaluation in LLMs},
  author = {Nataliya Stepanova and Ivan Titov and Emily Allaway and Björn Ross},
  year = {2026},
  eprint = {2610.08747},
  archivePrefix = {arXiv},
  primaryClass = {cs.CL},
  url = {https://arxiv.org/abs/2610.08747}, 
  }

• How broad is that claim? Mapping Generalisation in NLP Research
  Chenxin Diao, Nataliya Stepanova, Emily Allaway
  To appear at EMNLP 2026
  Paper Code

  BibTeX

  @misc{diao2026broadclaimmappinggeneralisation,
  title = {How broad is that claim? Mapping Generalisation in NLP Research},
  author = {Chenxin Diao and Nataliya Stepanova and Emily Allaway},
  year = {2026},
  eprint = {2609.14770},
  archivePrefix = {arXiv},
  primaryClass = {cs.CL},
  url = {https://arxiv.org/abs/2609.14770}
  }

• Concepts in terms of other Concepts: a method for the analysis of collective word use
  Gustavo Cilleruelo, Nataliya Stepanova, Emily Allaway, Alexandra Birch
  CogSci 2026
  Paper

  BibTeX

  @inproceedings{cilleruelo2026concepts,
  title = {Concepts in terms of other Concepts: a method for the analysis of collective word use},
  author = {Cilleruelo, Gustavo and Stepanova, Nataliya and Allaway, Emily and Birch, Alexandra},
  booktitle = {Proceedings of the Annual Meeting of the Cognitive Science Society},
  volume = {48},
  year = {2026}
  }

• Temporal generalizability in multimodal misinformation detection
  Nataliya Stepanova, Björn Ross
  GenBench Workshop at EMNLP 2023
  Paper

  BibTeX

  @inproceedings{stepanova2023temporal,
  title = {Temporal generalizability in multimodal misinformation detection},
  author = {Stepanova, Nataliya and Ross, Björn},
  booktitle = {Proceedings of the 1st GenBench Workshop on (Benchmarking) Generalisation in NLP},
  pages = {76--88},
  year = {2023}
  }

Education

• PhD Candidate, University of Edinburgh Sept. 2025 - present With the School of Informatics (Institute of Language, Cognition and Computation).
• MSc Speech and Language Processing Sept. 2021 – Aug. 2022 Funded by the Marshall Scholarship, graduated with Distinction.
• BSc Mathematics, BSc Computer Science Aug. 2017 – May 2021 Minored in Linguistics, funded by the Banneker-Key full merit scholarship, graduated Magna Cum Laude with a GPA of 3.98.

Professional Experience

• Software Engineer, Amazon Alexa (Cambridge, UK) Oct. 2022 – May 2024 Worked on integrating, optimizing, and deploying trained Large Language and Speech models into production environment.
• Software Engineering Intern, Capital One Financial Corporation (McLean, Virginia, USA) June 2021 – Aug. 2021 Worked on an ML-based SlackBot to automate question answering in Slack support channels.
• Software Engineering Intern, Capital One Financial Corporation (Plano, Texas, USA) June 2020 – Aug. 2020 Designed algorithm using NLP (sentiment extraction) to process agent/customer call and chat interactions to create a “smart note” describing the status of the customer’s refinancing application.

Teaching

Teaching assistant roles:
· Text Technologies for Data science (Fall 2025 - present)
· Accelerated Topics in NLP (Spring 2026)
· Computational Social Science (Spring 2026)
· Discrete Mathematics and Probability (Fall 2025)