Amir Saeidi

I am a fourth year PhD student in CS at Arizona State University, advised by Chitta Baral in the Cognition and Intelligence Lab. I am also a Research Intern at Microsoft, where I work on agentic AI and tool-use learning. My research focuses on aligning and improving large language models through preference optimization, building reliable tool-using agents, and evaluating the robustness of LLMs.

Some of my representative works include: VULCAN represents my work in the agentic AI domain, where agents learn to use tools by living in simulated environments (done at Microsoft); TPO represents my work in LLM alignment, introducing a triple preference optimization approach; DCPO represents my work in text-to-image alignment, applying preference optimization to diffusion models; and IRMA represents my work in tool-use reliability, improving accuracy in complex dynamic environments.

Representative papers are highlighted.

This website's source code is borrowed from Jon Barron. Much thanks!