Hi there 👋 I'm Eunbeen Hong

I'm an M.S./Ph.D. integrated student at CVLAB, KAIST AI, advised by Prof. Seungryong Kim. My research focuses on 4D scene understanding and generation, feed-forward novel view synthesis, and diffusion models.

---

🔬 Research

• FL4N: Feed-forward Latent Model for 4D Novel View Synthesis (arXiv 2026)
  • The first feed-forward latent 4D NVS model: renders a dynamic scene at unseen viewpoints and timestamps from a single monocular video, learning temporal dynamics as residuals on a frozen static latent prior [Paper] [Project] [Code]
• TETO: Tracking Events with Teacher Observation for Motion Estimation and Frame Interpolation (ECCV 2026)
  • Teacher-student framework that learns event-camera motion estimation from ~25 min of unannotated real data, then uses the motion to condition a video diffusion transformer for frame interpolation [Paper] [Project]
• Seg4Diff: Unveiling Open-Vocabulary Segmentation in Text-to-Image Diffusion Transformers (NeurIPS 2025) — analyzes joint image-text attention in MM-DiT and finds a "semantic grounding expert" layer whose attention maps already yield open-vocabulary segmentation masks; a lightweight LoRA fine-tuning on mask-annotated data (MAGNET) improves both segmentation and generation quality [Paper] [Project]
• S4M: Boosting Semi-Supervised Instance Segmentation with SAM (ICCV 2025)
  • Integrates SAM into a teacher-student framework via structural distillation, SAM-based pseudo-label refinement, and an instance-aware augmentation, reaching state-of-the-art results on semi-supervised instance segmentation benchmarks [Paper] [Project]

🛠️ Projects

• 🤖 LCBot — retrieval-augmented chatbot on an open-source backbone, built with Dell Technologies
• 🏆 TuitionShock.us — heatmap of U.S. college costs · 1st place, Justice Hacks @ CruzHacks 2024
• 🖼️ 3D: Diffusion Deception Detector — detects diffusion-generated images using PRNU noise residuals + ViT (>0.9 acc. on LSUN Bedroom)
• 🎵 LyriKOR — singable English→Korean lyric translation with syllable constraints · HCLT 2023, top 10 class project
• 🎮 Ola-Ola — Unity stair-climbing game with university rankings (Madcamp 2023)
• 👄 SeeSay — helping the deaf to speak

🎓 Education

• KAIST — M.S./Ph.D. Integrated, Kim Jaechul Graduate School of AI (2025 - Present)
• UC Santa Cruz — Exchange Student, Computer Science (2023 - 2024)
• Korea University — B.E. in Computer Science and Engineering (2021 - 2025), full-ride Academic Excellence Scholarship

---

💻 Tech Stack