Profile picture of Yana Veitsman

Hi! I’m a first-year PhD student at Universität Göttingen with Prof. Dr. Lisa Beinborn, working on cognitively plausible strategies for sample-efficient modeling in multilingual NLP.

My research interests revolve around sample-efficient learning, multilinguality, and interpretability.

In my PhD, I am tackling three questions:

  1. Which cognitively inspired inductive biases help language models learn from less data?
  2. Are these biases language-agnostic, or do they need to be tailored to specific languages?
  3. How can we build these biases into model architectures in a way that stays interpretable?

I completed my Master’s degree at Saarland University, where I worked as a student research assistant in the labs of Prof. Dr. Michael Hahn and Prof. Dr. Vera Demberg.

During my Master’s, I worked on defining the expressivity of state-space models, mechanistic interpretability of in-context learning, and persistent theoretical limitations of the transformer architecture.


Outside of work, I make other people laugh, lift heavy things, and experiment with the sourness level of my signature cocktails. In the past, I have held my solo stand-up comedy show. So, either see me performing or hit me up to discuss what rep range is the best for muscle hypertrophy over a Moscow Mule!