Junyoung Kim

email / cv / google scholar / github / linkedin

I received my Master’s degree from the Data Intelligence and Learning Lab (DIAL Lab) at Sungkyunkwan University (SKKU), Korea. My research explores how AI systems can understand and act on human intentions expressed through language, ultimately enhancing human–AI interaction.

Modeling Human Intent and Preference [1, 4, 5] — understanding implicit intentions and contextual goals in natural language to align system interpretation with human expression.
Human-centered Evaluation [3] — evaluation frameworks that move beyond static ground truths to capture user satisfaction, adaptability, and alignment with diverse human values.

My broader interest lies in using natural language to lower the barriers between people and information, building tools like information retrieval systems, search engines, and recommender systems that make knowledge easier to discover and use.

Outside of research, I enjoy reading, going for walks, and occasionally scribbling on my blog. Please feel free to reach out if you’re interested in my research!

Publications

[1] Bayesian Active Learning with Gaussian Processes Guided by LLM Relevance Scoring for Dense Passage Retrieval,
Junyoung Kim, Anton Korikov, Jiazhou Liang, Justin Cui, Yifan Liu, Qianfeng Wen, Mark Zhao, Scott Sanner,
ACL Findings, 2026 [arXiv] [code]
[2] Multimodal Item Scoring for Natural Language Recommendation via Gaussian Process Regression with LLM Relevance Judgments,
Yifan Liu, Qianfeng Wen, Jiazhou Liang, Mark Zhao, Justin Cui, Anton Korikov, Armin Toroghi, Junyoung Kim, Scott Sanner,
ACL Findings, 2026 [arXiv]
[3] Transparent Reference-free Automated Evaluation of Open-Ended User Survey Responses,
Subin An*, Yugyeong Ji*, Junyoung Kim*, Heejin Kook*, Yang Lu, Josh Seltzer,
EMNLP, Industry Track, 2025 [arXiv]
[4] Empowering Retrieval-based Conversational Recommendation with Contrasting User Preferences,
Heejin Kook*, Junyoung Kim*, Sungmin Park, Jongwuk Lee,
NAACL, 2025 [arXiv] [code]
[5] MARS: Matching Attribute-aware Representations for Text-based Sequential Recommendation,
Hyunsoo Kim*, Junyoung Kim*, Minjin Choi, Sunkyung Lee, Jongwuk Lee,
CIKM, Short Paper, 2024 [arXiv] [code]

* denotes equal contribution.

Experience

NAVER Corp. Jul 2026 – Oct 2026

AI Engineering Intern

Developed a self-improving LLM-as-a-Judge system for search-data annotation.
Developed an LLM-based system for translating natural-language label modification requests into executable code.

Angle Company Jan 2026 – Mar 2026

AI Research Intern

Developed a handwriting-separation and preprocessing pipeline for OCR on insurance documents, reducing Character Error Rate (CER) by up to 7%.

Data-Driven Decision-Making (D3M) Lab, University of Toronto Jan 2025 – Aug 2025

Research Assistant, advised by Prof. Scott Sanner

Led research on cost-aware document relevance modeling under LLM inference constraints, improving Recall@50 by up to 42.22% over existing methods [1].
Contributed to research on LLM-based relevance modeling for natural-language recommendation [2].

Nexxt Intelligence Jan 2025 – Jul 2025

Student Researcher (Industry–Academia Collaboration, Ministry of Science and ICT)

Developed an LLM-as-a-Judge framework for evaluating open-ended survey responses, achieving up to 0.8614 Spearman correlation with human experts [3].

Data Intelligence and Learning (DIAL) Lab, Sungkyunkwan University Apr 2022 – Jan 2025

Research Assistant, advised by Prof. Jongwuk Lee

Led research on LLM-based conversational recommendation with contrastive user preference modeling, improving Recall@10 by up to 99.72% over existing models [4].
Led research on attribute-aware text-based sequential recommendation, improving NDCG@10 by up to 29.26% (13%+ in zero-shot settings) over existing methods [5].

Education

Sungkyunkwan University, South Korea Sep 2023 – Aug 2025

M.S. in Artificial Intelligence

GPA: 4.36 / 4.5

Sungkyunkwan University, South Korea Mar 2019 – Aug 2023

B.S. in Computer Education

GPA: 4.17 / 4.5, Magna Cum Laude

Based on a design by Jon Barron.
Last updated by Junyoung Kim, Sep 2026.