About Me
Hi! I’m Dayoon Ko 😊, a Ph.D. candidate in Computer Science and Engineering at SNU, advised by Gunhee Kim. Currently, I’m a visiting researcher at UC Berkeley, working with Sewon Min.
I love to define practical problems that appear when people use AI in the real world. Recently, I’ve been focusing on on-device multimodal retrieval, searching what people see on their devices under tight memory and compute budgets. Before that, I asked whether search agents verify their claims and find the papers researchers need, how LLMs and RAG systems keep up with evolving knowledge, and whether models understand short-form video humor, an extreme case of implicit multimodal understanding.
Also, I feel excited when I finally find out how to solve the problems! For instance, scaling search agents with parallel and sequential reasoning or explicitly tracking which constraints an agent has verified, and detecting when a retriever needs an update.
Outside of research, I enjoy dancing 💃 or doing CrossFit 🏋🏻♀️. Staying active keeps my brain happy!
I love to define practical problems that appear when people use AI in the real world. Recently, I’ve been focusing on on-device multimodal retrieval, searching what people see on their devices under tight memory and compute budgets. Before that, I asked whether search agents verify their claims and find the papers researchers need, how LLMs and RAG systems keep up with evolving knowledge, and whether models understand short-form video humor, an extreme case of implicit multimodal understanding.
Also, I feel excited when I finally find out how to solve the problems! For instance, scaling search agents with parallel and sequential reasoning or explicitly tracking which constraints an agent has verified, and detecting when a retriever needs an update.
Outside of research, I enjoy dancing 💃 or doing CrossFit 🏋🏻♀️. Staying active keeps my brain happy!
Research Map
🔥 Recent News
[Oct 2026] Joined Sewon Min's group at UC Berkeley 🐻 as a visiting researcher!
[Oct 2026] New paper release! "CompLit": a scientific literature search benchmark for implicit, cumulative, and unmet needs.
[Oct 2026] New paper release! "ScholarCatalyst": a benchmark for finding papers that inspire new research.
[Sep 2026] Received the Yulchon AI Star Fellowship! 🎉
Publications
CompLit: Scientific Literature Search Benchmarks Must Cover Implicit, Cumulative, and Unmet Needs
arXiv 2026
Hybrid Deep Searcher: Scalable Parallel and Sequential Search Reasoning
ICLR 2026
Experiences
Visiting Researcher
UC Berkeley
Hosted by Sewon Min
October 2026 - Present
Research Intern
LG AI Research, Superintelligence Lab
Worked on agentic reasoning and retrieval systems for large language models
March 2025 - September 2026
Education
M.S./Ph.D. in Computer Science and Engineering
Seoul National University
Advisor: Gunhee Kim
September 2022 - Present
B.S. in Computer Science and Engineering
Yonsei University
GPA: 4.12/4.30 (Rank: 1/28)
March 2018 - August 2022
Honors
Yulchon AI Star Fellowship
Yulchon Foundation and Seoul National University AI Institute
2026
Graduation Exhibition Excellence Award
Yonsei University, Computer Science
Project on mitigating data imbalance in image classification
Feb 2022
Academic Excellence Award
Yonsei University
For semester grades
Fall 2021
Udeok Scholarship
Udeok Foundation
Academic excellence scholarship, with full tuition during undergraduate years
Feb 2019 - Aug 2022