Marlene is a last-year PhD Student at University of Mannheim, supervised by Markus Strohmaier. She works at the intersection of computational social science and responsible NLP, studying how social biases manifest across the model pipeline, from pre-training data and model internals to prompting strategies. She is particularly interested in how LLMs represent (and sometimes under- and misrepresent) different socio-demographic groups. Currently, she investigates how demographic patterns in the pre-training data connect to model outputs, with the goal of developing language technologies that are more diverse and socially responsible.