About Me

I’m MinJu Jeon, an AI researcher working on data-centric training for multimodal and language models. I’m currently a research intern at LG AI Research (EXAONE Lab).

My work explores how data curation, synthetic data generation, retrieval augmentation, and training-time data design shape what a model can do and how reliably it does it. In practice that spans multimodal understanding, dense video captioning, text-video retrieval, multilingual language and speech modeling, and agentic visual-code environments.

Before joining LG, I worked on multilingual speech (grapheme-to-phoneme, robust TTS) at NAVER Cloud Voice Tech, and on vision-language understanding at Hanyang University’s Multimodal AI Lab, advised by Prof. Dong-Jin Kim, where I completed my M.S. in Data Science in August 2026.

News

Aug 2026UGTPhon accepted at EMNLP 2026 (Findings, Long)

Aug 2026Completed M.S. in Data Science at Hanyang University

Jun 2026Joined LG AI Research, EXAONE Lab as a research intern

Mar 2026Cap4Bridge accepted at IEEE Access

Feb 2026Two papers accepted at CVPR 2026

Dec 2025Joined NAVER Cloud, Voice Tech as a research intern

Aug 2025Sali4Vid accepted at EMNLP 2025 (Long, Main)

Research Interests

Model Training & Data-Centric AI

pre-training mid-training post-training data curation synthetic data data mixture design retrieval augmentation robustness

Multimodal & Vision-Language

multimodal understanding vision-language models video-language learning document understanding GUI understanding

Agentic AI & Code Generation

visual code generation GUI agents tool use execution feedback trajectory data environment interaction

Experience

Research Intern, EXAONE Lab Multimodal foundation-model training, visual code generation, and synthetic trajectory data

Jun 2026 – Present

Research Intern, Voice Tech Multilingual G2P and robust TTS for non-canonical text · First-author paper at EMNLP 2026 Findings

Dec 2025 – Jun 2026

Research Assistant, Multimodal AI Lab Dense video captioning, text-video retrieval, and retrieval-augmented learning · Advisor: Dong-Jin Kim

Sep 2024 – Aug 2026

Selected Publications

EMNLP 2026 Findings Phonemizing User-Generated Text: A Benchmark, Taxonomy, and Compositional Approach EMNLP 2025 Main Sali4Vid: Saliency-Aware Video Reweighting and Adaptive Caption Retrieval for Dense Video Captioning CVPR 2026 Follow the Saliency: Supervised Saliency for Retrieval-augmented Dense Video Captioning IEEE Access 2026 Cap4Bridge: Caption-Guided Cross-Modal Contextualization with Stochastic Augmentation for Text-Video Retrieval CVPR 2026 SAIL: Similarity-Aware Guidance and Inter-Caption Augmentation-based Learning for Weakly-Supervised Dense Video Captioning ACM MM 2025 SynC: Synthetic Image Caption Dataset Refinement with One-to-many Mapping for Zero-shot Image Captioning

View all publications →