Sejoon Jun

I am a 4th-year undergraduate student at KAIST. My research interests lie in Generative&World models and Reasoning&Planning, and I am broadly excited about building intelligent systems that can perceive, predict, and interact with the physical world.

Currently, I am a Visiting Researcher at Northeastern University in Boston, working under the supervision of Prof. Lorenzo Torresani. At the same time, I am a research collaborator with Prof. Yilun Du at Harvard, actively working with Dr. Weirui Ye at MIT.

Sejoon Jun

News

Oct 2026
Sep 2026
May 2026
Our paper "Temporal Backtracking Search for Test-time Generative Video Reasoning" was accepted at the ICML 2026 FoGen Workshop.
Jan 2026
Started a research visit at Northeastern University under the supervision of Prof. Lorenzo Torresani.
Sep 2025
Our paper "Training-Free Safe Text Embedding Guidance for Text-to-Image Diffusion Models" was accepted at NeurIPS 2025.

Publications

C: conference papers. P: preprints and papers under review. Representative papers are highlighted.

MFM's memory evolution in Minecraft over 36,000 steps
TrajPilot teaser
Base model, training-based, previous training-free, and our training-free safety approaches for text-to-image diffusion

C1 Training-Free Safe Text Embedding Guidance for Text-to-Image Diffusion Models

Byeonghu Na, Mina Kang, Jiseok Kwak, Minsang Park, Jiwoo Shin, Sejoon Jun, Gayoung Lee, Jin-Hwa Kim, Il-Chul Moon

NeurIPS 2025

Pipeline for building a scene graph from image and point cloud observations for VLM reasoning

P1 Scene Graph Grounded Spatial Understanding with Vision-Language Models for Embodied Robotic Systems

Taewoo Kim, Kangje Moon, …, Sejoon Jun, …, Il-Chul Moon, Byeonghu Na

Under review