My Plan for my PhD

I plan to soon work on controlling text generation models to satisfy certain constraints like a given graph structure or a given linguistic structure for the generations. Thus, for the remaining years of my Ph.D., I want to work on the alignment and controllability of LLM generation. I am also interested in developing security protocols that allow us to distinguish human and machine-generated texts without sharing the model parameters. This allows to increase the reliability of these models by reducing their potential misuse due to easy and secure checking. In the future, I also plan to extend my experience in other types of structured data, particularly, relational tabular databases, and documents, due to their wide availability in the real-world infrastructure. My experience in general graph structures can prove extremely useful for my planned investigations into these structures.