Beatrix M. G. Nielsen

Beatrix M. G. Nielsen is a postdoc at the IT University of Copenhagen interested in understanding the representations of machine learning models. She has worked on figuring out how clustering works in high dimensions and what distances between probability distributions of models can tell us about how similar their representations are. Before she joined ITU, she did a bachelor's and master's in mathematics at the University of Copenhagen, spent some time in industry as a software developer and data scientist and did a PhD at the Technical University of Denmark.

References

  1. prediction_hubs_freq.png
    Prediction hubs are context-informed frequent tokens in LLMs
    Beatrix M. G. Nielsen, Iuri Macocco, and Marco Baroni
    In Proceedings of the 63nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), 2025
    URL
  2. distribution_space_to_rep.png
    When Does Closeness in Distribution Imply Representational Similarity? An Identifiability Perspective
    Beatrix M. G. Nielsen, Emanuele Marconato, Andrea Dittadi, and Luigi Gresele
    arXiv preprint arXiv:2506.03784, 2025
    URL
  3. Challenges in explaining representational similarity through identifiability
    Beatrix M. G. Nielsen, Luigi Gresele, and Andrea Dittadi
    In UniReps: 2nd Edition of the Workshop on Unifying Representations in Neural Models, 2024
  4. DiffEnc: Variational Diffusion with a Learned Encoder
    Beatrix M. G. Nielsen, Anders Christensen, Andrea Dittadi, and Ole Winther
    In The Twelfth International Conference on Learning Representations, 2024
  5. Hubness Reduction Improves Sentence-BERT Semantic Spaces
    Beatrix M. G. Nielsen and Lars Kai Hansen
    In Northern Lights Deep Learning Conference, 2024