Andrew Wan Ju Kang

Bio

I am a PhD graduate advised by Professor Se-Young Yun at the Kim Jaechul Graduate School of Artificial Intelligence, KAIST. My research interests span natural language generation and evaluation, particularly with vision-language models and the interplay between the different modalities within. Exploring the job market (academia and industry alike) has been my top priority in the recent months, so feel free to connect with me and have a chat!

Education

• PhD, KAIST
  Kim Jaechul Graduate School of Artificial Intelligence, 2025
  Thesis: Evaluating Vision-Language Representations towards Bias Measurement
  Advisor: Professor Se-Young Yun
  
• BS, KAIST
  School of Electrical Engineering, 2017
  Double Major in Business Technology Management
  

Research

• MondegreensEval: A Phonetic Benchmark for Measuring Language-model Bias in Automatic Speech Recognition.
  Wan Ju Kang
  ICML Workshop on ML For Audio, 2026. [Paper | Code | Dataset & Poster (coming soon) ]
• Benchmarking and Mitigating Modality Bias in Vision-language Models.
  Wan Ju Kang, Sumyeong Ahn, and Se-Young Yun.
  under review, 2026.
• Differential Diffusivity: Latent Diffusion Models as Zero-shot Evaluators of Image Descriptions.
  Wan Ju Kang, Ki Hoon Kwak, Haemin Choi, and James Thorne.
  under review, 2026.
• Are Large Vision-Language Models Ready to Guide Blind and Low-Vision Individuals?
  Na Min An, Eunki Kim, Wan Ju Kang, Sangryul Kim, Hyunjung Shim, and James Thorne.
  arXiv preprint, under review, 2026. [Paper | Project]
• Sightation Counts: Leveraging Sighted User Feedback in Building a BLV-aligned Dataset of Diagram Descriptions.
  Wan Ju Kang, Eunki Kim, Na Min An, Sangryul Kim, Haemin Choi, Ki Hoon Kwak, and James Thorne.
  ACL (Main), 2025. [Paper | Dataset | Poster]
• Political Slant Estimation and Mitigation with LLMs: A Parallactical Study on News Articles and Video News Transcripts.
  Wan Ju Kang, Jiyoung Han, Jaemin Jung, and James Thorne.
  working paper, 2025.
• Sampling from Bounded Clusters for Diverse Relevant Argument Retrieval.
  Wan Ju Kang, Jiyoung Han, Jaemin Jung, and James Thorne.
  ACL Workshop on Argument Mining, 2024. [Paper | Poster]

• Curiosity-driven multi-agent exploration with mixed objectives.
  Roben Delos Reyes, Kyunghwan Son, Jinhwan Jung Wan Ju Kang, and Yung Yi.
  arXiv preprint, 2022. [Paper]
• Reflect, not Regret: Understanding Regretful Smartphone Use with App Feature-Level Analysis.
  Hyunsung Cho, DaEun Choi, Donghwi Kim, Wan Ju Kang, Eun Kyoung Choe, and Sung-Ju Lee.
  ACM CSCW, 2021. [Paper]
• Inducing cooperation through reward reshaping based on peer evaluations in deep multi-agent reinforcement learning.
  David Earl Hostallero, Daewoo Kim, Sangwoo Moon, Kyunghwan Son, Wan Ju Kang, and Yung Yi.
  AAMAS, 2020. [Paper]
• Deep learning approaches to detect atrial fibrillation using photoplethysmographic signals: algorithms development study.
  Soonil Kwon, Joonki Hong, Eue-Keun Choi, Euijae Lee, David Earl Hostallero, Wan Ju Kang, Byunghwan Lee, Eui-Rim Jeong, Bon-Kwon Koo, Seil Oh and Yung Yi.
  JMIR mHealth and uHealth, 2019. [Paper]
• QTRAN: Learning to factorize with transformation for cooperative multi-agent reinforcement learning.
  Kyunghwan Son, Daewoo Kim, Wan Ju Kang, David Earl Hostallero and Yung Yi.
  ICML, 2019. [Paper]
• Learning to schedule communication in multi-agent reinforcement learning.
  Daewoo Kim, Sangwoo Moon, David Earl Hostallero, Wan Ju Kang, Taeyoung Lee, Kyunghwan Son, and Yung Yi.
  ICLR, 2019. [Paper | Slides]

Work & Teaching Experience

• NLP Specialist, NextUs (2023/07 - 2023/09)
  News curation service with political slant detection and measurement. Funded by the Korea Press Foundation.
• NLP Analyst, Seoul National University Asia Center (2022/01 - 2022/07)
  Political polarization trends in traditional and new media. Funded by the Seoul National University Asia Center.
• NLP and Generative AI Course Instructor
  Samsung Electronics (2024/02 - Present),
  National Intelligence Service (2026/07 - Present),
  KT (2025/06 - 2025/07),
  KAIST IT Academy (2023/01 - 2023/02),
  Daejeon Statistical Training Institute (2022/08),
  Incheon International Airport Academy (2022/08)
• Machine Learning Tutor
  Hyundai Motors (2023/11 - 2023/12),
  LG Chem (2021/12),
  SK Hynix (2019/07),
  Korea Institute of Communications and Information Sciences (2019/02)
• KAIST EE213 Discrete Methods for Electrical Engineering
  Head Teaching Assistant (2020/09 - 2020/12)
  
• KAIST CC500 Scientific Writing and HSS581 Graduate English Writing
  Teaching Assistant (2015/03 - 2017/12)
  
• The KAIST Herald (2012/09 - 2018/06)
  Editor-in-Chief

Honors and Awards

• Cohere Labs Catalyst Grant
  2025 Fall Cohort Recipient for Modality Utilization Imbalance in Multimodal Models (2025)
  
• Seocho AI College Grand Prize
  NLP-based Real-time Voice Phishing Detection in Phone Calls (2022)
  
• Samsung Humantech Award Bronze Prize
  Learning to Schedule Communication in Multi-agent Reinforcement Learning (2019)