Pietro Lesci

Pietro Lesci is a PhD student in the Department of Computer Science and Technology at the University of Cambridge. His research focuses on Machine Learning, Natural Language Processing (NLP), and Active Learning. Pietro's work addresses challenges such as tokenization bias, model stability, and efficient active learning for large datasets. He leads projects like ByteSpan (subword tokenization) and AnchorAL (active learning in imbalanced data).

Key research areas include causal estimation of memorization profiles, dialogue state tracking via Diable, and small language model convergence. His contributions span theoretical frameworks and applied techniques for improving NLP and machine learning systems.

Outside research, Pietro collaborates with the Cambridge Computational Linguistics group and contributes to open-source tools for language model analysis.