Portrait of Jinho Kang

Jinho Kang

Ph.D. student in Computer Science
The University of Texas at Dallas

Robust to

CorruptionNoise and blur at test time.
Domain shiftTrained on color drawings, tested on pencil sketches.
Label shiftThe class mix changes after training.
New classesObjects never seen in training.

About

My research focuses on building
reliable and robust artificial intelligence.

I currently study how Vision-Language Models (VLMs) generalize and adapt in open-world, non-stationary, and distributionally shifting environments.

I am a Ph.D. student in Computer Science at the University of Texas at Dallas, advised by Prof. Yunhui Guo. Prior to my Ph.D., I received my M.S. in Artificial Intelligence and B.S. in Computer Science from the University of Seoul. During my M.S., I was advised by Prof. Kyungwoo Song at Yonsei University and Prof. Jiyoung Jung at the University of Seoul.

10
Papers
9 published, 1 under review
6
First-author papers
first or co-first author
5+
Years in research
since joining a lab in 2021
4
Institutions
UT Dallas, CMU, Yonsei, Univ. of Seoul

Now ongoing projects

  • Since Sep 2026

    Open-Set Test-Time Adaptation of Vision–Language Models for Object Detection

    Source-free adaptation for reliable detection under distribution shift, preserving known-class recognition while localizing unfamiliar objects and explicitly distinguishing them from known categories and background.

  • Since Oct 2026

    Continual Learning for Long-Sequence Video QA with Hybrid Mamba

    Robust continual learning for long-sequence video QA, using the context-retention capabilities of hybrid Mamba architectures to preserve long-range temporal dependencies and mitigate forgetting across tasks.

  • Since Aug 2026

    Psychological Memory for Long-Term Conversational Agents

    Agentic construction of human-grounded datasets and benchmarks to evaluate how long-term conversational agents integrate individual traits and evolving psychological states into personalized, supportive responses.

Publications

In review

  • Preprint

    CAMA: Calibrated Anti-Collapse Marginal Alignment for Test-Time Adaptation of Vision–Language Models

    Jinho Kang, Yunhui Guo

    Submitted to AI Top Conference 2027

2026

  • SAC 2026Conference

    CBPE: Causal-aware Blood Pressure Estimation with Multimodal Data

    Hoyoon Byun, Sumin Park, Jinho Kang, Hyeonjeong Lee, Minseong Kim, Doosik Kim, Kwang-Yong Kim, Kyungwoo Song

    ACM Symposium on Applied Computing (SAC), 2026

    Paper Code

2025

  • ICDM 2025Conference, short paper

    Towards Generalizable Time Series Forecasting via IB-Regularized Transformer-Diffusion

    Dagyeong Na, Jinho Kang, Junseok Kwon

    IEEE International Conference on Data Mining (ICDM), 2025

    Paper
  • CIKM 2025Conference

    Causal Effect Variational Transformer for Public Health Measures and COVID-19 Infection Cluster Analysis

    Jinho Kang, Sungjun Lim, Hojun Park, Jiyoung Jung, Jaehun Jung, Kyungwoo Song

    ACM International Conference on Information and Knowledge Management (CIKM), 2025

    Paper Code
  • IEEE Access 2025Journal

    Enhancing Global and Local Context Modeling in Time Series Through Multi-Step Transformer-Diffusion Interaction

    Dagyeong Na, Jinho Kang, Byoungwoo Kang, Junseok Kwon

    IEEE Access, 2025 SCIE, JCR 2024 IF=3.9

    Paper
  • KBS 2025Journal

    Multi-Query Frequency Prompting for Physiological Signal Domain Adaptation

    Jinho Kang*, Hoyoon Byun*, Taero Kim, Jiyoung Jung, Kyungwoo Song

    Knowledge-Based Systems, 2025 SCIE, JCR 2024 IF=7.6

    Paper Code
  • UAI 2025Conference

    Flat Posterior Does Matter For Bayesian Model Averaging

    Sungjun Lim, Jeyoon Yeom, Sooyon Kim, Hoyoon Byun, Jinho Kang, Yohan Jung, Jiyoung Jung, Kyungwoo Song

    Conference on Uncertainty in Artificial Intelligence (UAI), 2025

    Paper arXiv

2024

  • IEEE Sensors J. 2024Journal

    Few-Shot PPG Signal Generation via Guided Diffusion Models

    Jinho Kang, Yongtaek Lim, Kyuhyung Kim, Hyeonjeong Lee, Kwang-Yong Kim, Minseong Kim, Jiyoung Jung, Kyungwoo Song

    IEEE Sensors Journal, 2024 SCIE, JCR 2022 IF=4.3

    Paper Code
  • IJCAI AI4TS 2024Workshop Best Paper Honorable Mention

    Sequential Treatment Effect Estimation with Variational Transformers: Application to COVID-19 Infection Clusters

    Jinho Kang, Sungjun Lim, Hojun Park, Jaehun Jung, Jiyoung Jung, Kyungwoo Song

    IJCAI Workshop on AI for Time Series Analysis (AI4TS), 2024

  • Pattern Recognition 2024Journal

    Bibimbap: Pre-trained models ensemble for Domain Generalization

    Jinho Kang, Taero Kim, Yewon Kim, Changdae Oh, Jiyoung Jung, Rakwoo Chang, Kyungwoo Song

    Pattern Recognition, 2024 SCIE, JCR 2022 IF=8.0

    Paper Code

* equal contribution

Background

Experience

  1. Aug 2025 – Dec 2025

    Associate Researcher

    Machine Learning and Artificial Intelligence Lab, Yonsei University

    Seoul, South Korea

  2. Aug 2024 – Feb 2025

    Visiting Scholar

    School of Computer Science, Carnegie Mellon University

    Pittsburgh, PA, USA. Supported by IITP (Institute for Information & Communication Technology Planning & Evaluation).

  3. Apr 2023 – Aug 2025

    Graduate Researcher

    Machine Learning and Artificial Intelligence Lab, University of Seoul

    Seoul, South Korea

  4. Sep 2021 – Apr 2023

    Undergraduate Researcher

    Machine Learning and Artificial Intelligence Lab, University of Seoul

    Seoul, South Korea

Education

  1. Jan 2026 – Dec 2030 (expected)

    Ph.D. in Computer Science

    The University of Texas at Dallas

    Richardson, TX, USA. Advisor: Yunhui Guo

  2. Mar 2023 – Aug 2025

    M.S. in Artificial Intelligence

    University of Seoul

    Seoul, South Korea. Advisor: Jiyoung Jung. Co-advisor: Kyungwoo Song (Yonsei University)

  3. Mar 2016 – Feb 2023

    B.S. in Computer Science

    University of Seoul

    Seoul, South Korea

Past projects

  • Feb 2025 – Aug 2025

    Failure-Aware Causal Modeling through FMEA-Driven LLM Adaptation

    Leveraging domain-specific technical documents from the automotive industry to pretrain and adapt LLMs for generating accurate, component-targeted FMEA scenarios via causal reasoning.

  • May 2024 – Oct 2024

    Causal Discovery in Photoplethysmography Attributes

    Uncovering causal relationships among PPG signals and diverse physiological attributes (e.g., race, age) to enable robust and accurate arterial blood pressure (ABP) estimation.

    Code
  • Feb 2024 – May 2024

    Conditioned Generation for Aircraft Trajectory Data

  • May 2023 – Feb 2024

    Domain Generalization via Synthetic Photoplethysmography Data Generation

    Address data imbalance in medical datasets by using regressor-guided diffusion models to generate high-quality, diverse PPG signals, significantly enhancing Arterial Blood Pressure predictions.

    Code
  • Mar 2023 – Feb 2024

    Causal Prediction of COVID-19 Infection Cluster Dynamics

    Developed a machine learning model using a Transformer to predict the infection spread and duration of coronavirus clusters by embedding and processing both cluster and patient information.

    Code
  • Jul 2022 – Feb 2023

    Robust Ensemble Learning for Chemical Graph Data

Awards and honors

  • Aug 2024

    Best Paper Honorable Mention

    IJCAI Workshop on AI for Time Series Analysis (AI4TS)

    For “Sequential Treatment Effect Estimation with Variational Transformers”.

  • Jul 2023

    Summer Academic Paper Presentation Conference

    Korean Data Analysis Society (KDAS)

    For “Bibimbap: Pre-trained models ensemble for Domain Generalization”.