I am a second-year PhD student at Seoul National University, working with Prof. Gunhee Kim. I am also a visiting researcher at Stanford University, hosted by Prof. Chelsea Finn.

I have always been drawn to what I know, what I do not, and how to explore the gap between them. This leads me to ask how models can use external knowledge to solve increasingly challenging problems. I believe we already have plenty of knowledge to draw on, especially the knowledge that keeps evolving and accumulating. We need memory systems that models can manage on their own, so they can reuse prior knowledge instead of reasoning from scratch and avoid repeating past failures.

Such knowledge is hard to retrieve with existing methods, since it rarely looks similar on the surface. Still, humans (especially domain experts) intuitively know where to look in a vast search space and find connections that seem obvious only in hindsight, a kind of research taste explored in ScholarCatalyst. I aim to build models that find such connections for better decision making in open-ended, long-horizon problems.

To learn more about how I think beyond research, read my Blog.

Publication

ScholarCatalyst: A Benchmark for Retrieving Papers that Inspire New Research

Sohyeon Kim^*, Yoonho Lee^*, Bo Liu, Dayoon Ko, Rulin Shao, Seungone Kim, Graham Neubig, Pang Wei Koh, Aakanksha Chowdhery, Akari Asai, Omar Khattab, Yejin Choi, Gunhee Kim, Chelsea Finn
Preprint 2026  Website  Paper  Code  Data  *equal contribution

ScholarCatalyst is a benchmark built from AI researchers’ firsthand accounts of what inspired their work. A paper that inspires a project may share little surface similarity with the question (right), while topically related work may not (left). Even strong retrieval systems, including agents, find only about half of such papers.

MULTI3IR: A Benchmark for Multi-perspective Multi-domain Multi-modal Information Retrieval

Seokwon Song, Sohyeon Kim, Gunhee Kim
EMNLP 2026  Paper

When Is Enough Not Enough? Illusory Completion in Search Agents

Dayoon Ko, Jihyuk Kim, Sohyeon Kim, Haeju Park, Dahyun Lee, Gunhee Kim, Moontae Lee, Kyungjae Lee
COLM 2026 Workshop on Agent Behavior  Paper

Hybrid Deep Searcher: Scalable Parallel and Sequential Search Reasoning

Dayoon Ko, Jihyuk Kim, Haeju Park, Sohyeon Kim, Dahyun Lee, Yongrae Jo, Gunhee Kim, Moontae Lee, Kyungjae Lee
ICLR 2026  Paper  Website

When Should Dense Retrievers Be Updated in Evolving Corpora? Detecting Out-of-Distribution Corpora Using GradNormIR

Dayoon Ko, Jinyoung Kim, Sohyeon Kim, Jinhyuk Kim, Jaehoon Lee, Seonghak Song, Minyoung Lee, Gunhee Kim
ACL 2025 Findings  Paper