Pandas

Hyangsuk Min is a PhD student at KAIST, advised by Prof. Hwanjun Song. Her research asks how AI agents stay trustworthy when the information they receive is incomplete, noisy, evolving, or spread across contexts — through data-centric evaluations of how models perceive, structure, remember, and use information, and through memory-enabled agents that reuse past interactions without losing faithfulness. She also studies AI co-scientist systems, testing whether LLMs show researcher-level understanding when reading, comparing, and synthesizing scientific literature.