Dilara Soylu is a PhD student in Computer Science at Stanford University, supported by the School of Engineering Ravi Family Graduate Fellowship. Dilara has been studying the capabilities, limitations and societal applications of large language models; most recently focusing on developing programmatic ways to steer model generations towards target properties. At the RegLab, Dilara works on finding better ways to make use of external context to increase large language models’ performance on reasoning tasks. When not injured, she enjoys going on walking tours with friends, attempting outdoor sports, teaching computer science and playback theater.