Marzena Karpinska

prospective students

I plan to take ~1 PhD student every year in the Fall. Please apply through the standard SFU process. There is no need to email me. If you would like to work with me on a project, fill in this form. SFU students can also join through CMPT 415/416. Please note that any AI-generated email will be deleted without consideration.

More details

Can you consult on my application or answer my email?

Sadly, no. Due to the volume of emails I receive, I cannot consult on applications or respond to individual emails from prospective students. I check the form periodically and will get back to you if there is a good fit.

Master's or PhD?

SFU, like other Canadian universities, offers separate Master's and PhD programs. I may need to prioritize PhD applications and accept Master's students only when funding permits. You can also apply directly to a PhD program.

What is CMPT 415/416?

It is a research course that lets you do a research project for credit over two semesters. You will meet with me weekly and report your progress. We work as a team, and while it is a lot of work, it is also a lot of fun. You can also help with research without taking the course if you are motivated and serious about it.

about me

I am currently an assistant professor at the Simon Fraser University in beautiful Vancouver, Canada. Before that, I was a senior researcher at Microsoft based in Redmond. I did my postdoc at the Manning College of Information & Computer Sciences, University of Massachusetts Amherst working with Prof. Mohit Iyyer.

I hold a Ph.D. from the Department of Language and Information Sciences at the University of Tokyo.

research

My research focuses on natural language processing (NLP) and language models (LMs). Specifically, I am interested in:

Detection of AI-generated and AI-translated text: what characterizes text that is generated or translated by AI, and how well can we distinguish it from human-authored text?
Multilingual NLP: how can we build systems that serve people from diverse linguistic and cultural backgrounds?
Machine translation of creative texts: how can we help more authors be heard without misrepresenting their voice?
Long-form text processing: how well can models process long-form input, and how can we evaluate long-form output?

media

mentions refers to our findings · discusses features our work · quotes quoted

EL PAÍS
Can one AI unmask another? Pangram, an almost infallible tool for detecting machine-created texts and novels quotes
RADIO-CANADA
Controverse Thélyson Orélien : ce qu’il faut savoir sur Pangram quotes
NATURE
AI-detection tools have made huge leaps forward — how good are they?quotes
FUTURISM
Study: New York Times Has Published Extensive AI-Generated Articles mentions
THE ATLANTIC
How AI Is Creeping Into The New York Times discusses
WSJ
An AI Upheaval Is Coming for Media. This Journalist Is Already All In mentions
PRESS GAZETTE
Study claims 9% of US newspaper articles at least partly AI generated discusses
FORBES
Study Shows Experienced Humans Can Spot Text Created By AI discusses
THE ECONOMIST
GPT, Claude, Llama? How to tell which AI model is best mentions
TECH CRUNCH
Gemini's data-analyzing abilities aren't as good as Google claims discusses
UMASS MAGAZINE
Literary Device discusses
SLATOR
In Literary Translation, Humans Prefer Humans and Machines Prefer Machines discusses

news

Aug 2026 Our paper on long-context literary machine translation was accepted to EMNLP Findings 🎉
Aug 2026 Our lab received a 5-year NSERC Discovery Grant & Launch Supplement 🎉
Apr 2026 Our paper on AI-generated text in US news articles was accepted to ACL 2026 (oral, 3.9% papers) 🎉
Jan 2026 Started as an assistant professor at Simon Fraser University in Vancouver, Canada.
Oct 2025 New paper on AI-generated text in US news articles is out
Aug 2025 Papers on interdisciplinary approach to MT, cross-lingual memorization, and quantization effects on long-context tasks accepted to EMNLP 2025 🎉
Jul 2025 Our work on multilingual long-context processing was accepted to COLM 🎉
Jun 2025 Preprint on interdisciplinary approach to machine translation is out
Jun 2025 Serving as a Senior Area Chair at EMNLP
May 2025 Our preprint on cross-lingual memorization is out
May 2025 Our preprint on the effect of quantization on long-context tasks is out
May 2025 Our works on AI-generated text and multilingual long-form QA were accepted to ACL 🎉
Mar 2025 Our preprint on multilingual long-context processing is out
Jan 2025 Our preprint on machine generated text detection is out
Jan 2025 New preprint on slow-down attacks on reasoning models is out
Nov 2024 Recognized as an outstanding AC at EMNLP 2024
Sep 2024 NoCha was accepted to EMNLP 2024 🎉
Sep 2024 ESA was accepted to WMT 2024 🎉
Aug 2024 Presented our work on the evaluation of long-context language models at UNSW
Jul 2024 Presented our work on the evaluation of long-context language models at RMIT
Jul 2024 Presented our work on the evaluation of long-context language models at the University of Melbourne
Jul 2024 FABLES was accepted to COLM 2024 🎉
Jun 2024 Our preprint on LONG-CONTEXT processing capabilities of language models is out
Jun 2024 Our preprint on MULTI-LINGUAL/CULTURAL performance of language models is out
Jun 2024 Our preprint on more robust evaluation for machine translation is out
Apr 2024 Our preprint on faithfulness in book-length summaries is out
Mar 2024 NarrativeTime was accepted to LREC-COLING 2024 🎉
Dec 2023 Presented our work at WMT in Singapore.
Nov 2023 Launched litmt.org, a platform for sharing machine-translated world literature.
May 2023 Virtual talk at Instituto Superior Técnico & Unbabel Seminar on translation with Large Language Models.
Apr 2023 Virtual talk at Microsoft MT Reading Group on translation with Large Language Models.
Jan 2023 Virtual talk at Polish Academy of Sciences on Evaluation of Long-form Text Generation.
Dec 2022 Presented our work on diagnosing automatic evaluation metrics at EMNLP in Abu Dhabi.
Dec 2022 Presented our work on document-level MT at EMNLP in Abu Dhabi.
Nov 2022 Slator on literary MT and our PAR3 work