Mir Tafseer Nayeem

Mir Tafseer Nayeem

User-centric AI researcher. Huawei PhD Fellow.

Mir Tafseer Nayeem

I study how human context shapes language models, their applications, and their evaluation.

People produce the language data that models learn from. Tokenization turns that data into units models use to predict what comes next. My research follows this path from who creates information to whom AI systems serve, and asks how we can evaluate the outcomes.

I am a final-year PhD Candidate in Computing Science at the University of Alberta, advised by Prof. Davood Rafiei.

Huawei PhD Fellow EMNLP 2024 Best Resource Paper COLING 2018 Area Chair Favorite IEEE VIS 2025 Best Short Paper

Applications

How can AI applications better serve diverse people?

People create and use information in different ways, shaped by their experience, preferences, language, culture, and age. I use that context to select evidence, design outputs, and evaluate whether applications serve their intended audiences.

Responsible Evaluation

What do benchmark scores miss about users and their contexts?

Benchmarks can hide whose data and judgments define success. I audit their source and target users, build multilingual evaluations of model judges, and test how country income and deceptive chart designs affect model outputs and users' interpretations.

Announcements

  1. Papers at EMNLP, AACL, and NeurIPS (Workshop).

  2. One paper accepted at ACL 2026.

  3. One paper accepted at EACL 2026.

  4. ChartDeception received the Best Short Paper Award at IEEE VIS 2025.

  5. KidLM received the Best Resource Paper Award at EMNLP 2024.