Yuekun Yao

About me

photo

Hi! I am Yuekun Yao, a Ph.D. student in Department of Language Science and Technology at Saarland University working with Prof. Alexander Koller. I am a part of Computational Linguistic Group. In the past, I got my MSc degree in artificial intelligence at the University of Edinburgh. Before that, I did my BS in computer science at East China Normal University.

The main research question I am interested in is How does NLP models generalize to unfamiliar data and how can we improve it? I investigate out-of-distribution generalization with a focus on compositional generalization to bridge the gap between training and test distributions in realistic applications. I am also interested in trustworthiness of NLP models to detect their generalization errors when deployed in real-world settings.

My work aims to both understand model behaviours and develop more effective and reliable NLP models through the following research questions.

  • Can NLP models perform human-like generalization, and why? [1] Does this also apply to large language models? [2]
  • How to improve models' compositional generalization ability with general-purpose models (seq2seq)? [3]
  • How to build trustworthy models that generalize reliably? Can we train one model (discriminator) to judge the outputs of another model (parser)? [4]

Publications [Google Scholar][Semantic Scholar]

2025

Anything Goes? A Crosslinguistic Study of (Im)possible Language Learning in LMs
Xiulin Yang, Tatsuya Aoyama, Yuekun Yao, Ethan Wilcox
Preprint, 2025
2024
Predicting generalization performance with correctness discriminators [paper]
Yuekun Yao, Alexander Koller
Findings of the 2024 Conference on Empirical Methods in Natural Language Processing (EMNLP 2024)
Simple and effective data augmentation for compositional generalization [paper] [code]
Yuekun Yao, Alexander Koller
The 2024 Annual Conference of the North American Chapter of the Association for Computational Linguistics (NAACL 2024)
2023
SLOG: A Structural Generalization Benchmark for Semantic Parsing [paper] [code]
Bingzhi Li, Lucia Donatelli, Alexander Koller, Tal Linzen, Yuekun Yao, Najoung Kim
The 2023 Conference on Empirical Methods in Natural Language Processing (EMNLP 2023)
2022
Structural generalization is hard for sequence-to-sequence models [paper] [code] [data]
Yuekun Yao, Alexander Koller
The 2022 Conference on Empirical Methods in Natural Language Processing (EMNLP 2022)
2020
Dynamic masking for improved stability in online spoken language translation [paper]
Yuekun Yao, Barry Haddow
The 14th Biennial Conference of the Association for Machine Translation in the Americas (AMTA 2020)
ELITR non-native speech translation at IWSLT 2020 [paper]
Dominik Macháček, Jonáš Kratochvíl, Sangeet Sagar, Matúš Žilinec, Ondřej Bojar, Thai-Son Nguyen, Felix Schneider, Philip Williams, Yuekun Yao
The 17th International Conference on Spoken Language Translation (IWSLT 2020)

Contact me

ykyao [dot] cs [at] gmail [dot] com