Junhyung Ahn

614 E. Daniel St.

Champaign, IL 61820

I am a Ph.D. student at the University of Illinois Urbana-Champaign (UIUC) School of Information Sciences, where I am fortunate to be advised by Prof. Yonghan Jung.

Previously, I was an AI Engineer at NAVER, where I worked on off-policy evaluation for large-scale online ad auction systems and designed an LLM-as-a-Judge evaluation framework for automated quality assessment of AI-generated search responses, including bias mitigation strategies and an ensemble judging pipeline to replace human annotation at scale. I received my M.S. and B.S. in Electrical Engineering from KAIST, advised by Prof. Changho Suh.

Research interests

Causal inference: counterfactual reasoning, off-policy evaluation
Agentic AI: LLM-as-a-judge, evaluation bias mitigation
Recommender systems: data bias and scarcity, cold-start problems

news

Aug 2026	Started my Ph.D. at the UIUC School of Information Sciences.
Nov 2025	Our paper on counterfactual post-click conversion rate prediction was accepted at IEEE ICDM 2025: Best Paper Award Finalist.
Aug 2025	Joined NAVER as an AI Engineer on the Causal Inference team.

selected publications

ICDM
On Predicting Post-Click Conversion Rate via Counterfactual Inference

Junhyung Ahn and Sanghack Lee

In IEEE International Conference on Data Mining (ICDM), 2025

Best Paper Finalist Abs DOI arXiv Bib HTML

Best Paper Award Finalist, IEEE ICDM 2025

Conversion rate prediction suffers from selection bias: conversion labels are observed only for clicked impressions. We cast the missing labels as a counterfactual quantity and impute conversions for non-clicked samples via counterfactual inference, improving both offline estimation and online conversion rate.
@inproceedings{ahn2025postclick, title = {On Predicting Post-Click Conversion Rate via Counterfactual Inference}, author = {Ahn, Junhyung and Lee, Sanghack}, booktitle = {IEEE International Conference on Data Mining (ICDM)}, pages = {11--20}, year = {2025}, doi = {10.1109/ICDM65498.2025.00008}, }

IEEE TIT

On the Fundamental Limits of Matrix Completion: Leveraging Hierarchical Similarity Graphs

Junhyung Ahn, Adel Elmahdy, Soheil Mohajer, and 1 more author

IEEE Transactions on Information Theory, 2024

DOI arXiv Bib HTML

@article{ahn2024fundamental,
  title = {On the Fundamental Limits of Matrix Completion: Leveraging Hierarchical Similarity Graphs},
  author = {Ahn, Junhyung and Elmahdy, Adel and Mohajer, Soheil and Suh, Changho},
  journal = {IEEE Transactions on Information Theory},
  volume = {70},
  number = {3},
  pages = {2039--2075},
  year = {2024},
  doi = {10.1109/TIT.2023.3345902},
}

NeurIPS
Matrix Completion with Hierarchical Graph Side Information

Adel Elmahdy^*, Junhyung Ahn^*, Changho Suh, and 1 more author

In Advances in Neural Information Processing Systems (NeurIPS), 2020

Equal contribution.

Abs arXiv Bib HTML PDF

We consider a matrix completion problem that exploits social or item similarity graphs as side information, characterize the information-theoretic limit on the number of observed entries needed for exact recovery, and give a computationally efficient algorithm that achieves it.
@inproceedings{elmahdy2020matrix, title = {Matrix Completion with Hierarchical Graph Side Information}, author = {Elmahdy, Adel and Ahn, Junhyung and Suh, Changho and Mohajer, Soheil}, booktitle = {Advances in Neural Information Processing Systems (NeurIPS)}, volume = {33}, pages = {9061--9074}, year = {2020}, note = {Equal contribution.} }