Members

I have always operated at the seams between disciplines. While my academic home is in Computer Science and NLP, my own background is multi-faceted, and I’ve spent years navigating the dialogue between formal logic and the nuanced, often subjective worlds of human thought and culture.

I founded the NLP for Human Sciences Lab at Ariel University to be a reflection of this journey. My work focuses on building the bridges needed to connect high-end AI with fields that were long considered "too complex" for computation. I thrive on translating "soft" human questions into rigorous computational models, and I believe machine learning and language models can play a pivotal role in reshaping science and advancing our understanding of the world, and particularly of ourselves.

Let’s connect: ayal dot s dot klein at gmail  

I am developing a novel framework for concept-based model control to address the challenges of transparency and safety in LLMs. Our methodology leverages low-rank subspaces and concept activation vectors to isolate specific cognitive signals within neural networks. By reducing the subspace of activation outputs, we provide a more interpretable lens into how knowledge is encoded, enabling more precise 'surgical' interventions in model outputs.