Polina Dolgova

Hi! I am a PhD student at CISPA in Germany, working with Sebastian U. Stich.

📐 Prior to my PhD, I studied both pure mathematics (during my bachelor’s at HSE in Moscow) and computer science (master’s at HSE, Yandex School of Data Analysis). This background strongly shaped my interest in mathematical explanations and theoretical motivations behind learning algorithms :)

🏭 I also spent several years in industry: working with LLMs at Yandex (russian bigtech) and on audio/CV problems in a pharma R&D team in Spain.

🧠 My research interests center on machine unlearning. I’m particularly interested in how to define and reason about forgetting in modern models, and in the analysis of unlearning algorithms. More broadly, I enjoy problems where mathematical structure helps clarify how learning algorithms behave.

research interests

• Machine unlearning: formal definitions, data preservation/removal, and data attribution
• Continual learning: methods for retaining useful knowledge under model updates
• Stability, efficiency and robustness of learning algorithms
• Alignment-related questions of data influence, controllable updates, and reliable model behavior
• Mathematical foundations of machine learning and theory-driven empirical methods