Wonjun Jo

I am a Ph.D. candidate in Electrical Engineering at POSTECH, advised by Prof. Tae-Hyun Oh at KAIST’s AMI Lab. I also conduct research with the Physical AI group at KIST, advised by Prof. In So Kweon.

My research focuses on embodied AI and robot learning, building on self-supervised and multimodal learning. Recently, I have been studying how robots can adapt to new tasks and acquire reusable skills, focusing on vision-language-action (VLA) model adaptation and agentic robot learning.

Embodied AI
Self-supervised Learning
Multi-modal Learning
Robot Learning

| CV | Scholar | GitHub | LinkedIn

Current Research

Comparison of behavior cloning fine-tuning with and without Base Action-Field Regularization for controlling policy drift

When Fine-Tuning Drifts: Mitigating Policy Drift with Base Action-Field Regularization
Wonjun Jo, Nam Hyeon-Woo, Yohan Park, Hyunwoo Ha, Tae-Hyun Oh
Under review, 2026
Project Page | Paper | arXiv

Mitigating policy drift for robust few-shot fine-tuning of pretrained robot policies.

One-shot Teaching for Reusable Agentic Robot Skills
Wonjun Jo
In progress
Project Page | Paper | arXiv

Peer-Reviewed Publications

LiDAR-anchored collaborative distillation: initial features, Stage 1 image-to-LiDAR alignment, and Stage 2 LiDAR-to-image adaptation

LiDAR-Anchored Collaborative Distillation for Robust 2D Representations
Wonjun Jo, Hyunwoo Ha, Kim Ji-Yeon, Hawook Jeong, Tae-Hyun Oh
IEEE Robotics and Automation Letters (RA-L), 2026
Project Page | Paper | arXiv

Improving the robustness and generalization of self-supervised visual representations with LiDAR guidance.

DarkQA low-light visual question answering benchmark overview

DarkQA: Benchmarking Vision-Language Models on Visual-Primitive Question Answering in Low-Light Indoor Scenes
Yohan Park, Hyunwoo Ha, Wonjun Jo, Tae-Hyun Oh
IEEE Robotics and Automation Letters (RA-L), 2026
Project Page | Paper | arXiv

Benchmarking the robustness of embodied vision-language models under low-light conditions.

Collaborative distillation method for lighting-robust and 3D-aware representations

Self-Supervised Collaborative Distillation: Enhancing Lighting Robustness and 3D Awareness
Wonjun Jo, Hyunwoo Ha, Kim Ji-Yeon, Hawook Jeong, Tae-Hyun Oh
Workshop on Wild3D, IEEE/CVF International Conference on Computer Vision (ICCV), 2025
Paper | OpenReview

Improving pretrained visual representations for lighting robustness and 3D-aware generalization.

Overview of simple remedies for image-to-LiDAR representation learning

The Devil is in the Details: Simple Remedies for Image-to-LiDAR Representation Learning
Wonjun Jo, Kwon Byung-Ki, Kim Ji-Yeon, Hawook Jeong, Kyungdon Joo, Tae-Hyun Oh
Asian Conference on Computer Vision (ACCV), 2024
Project Page | Paper | arXiv

Improving self-supervised image-to-LiDAR representation learning for robust 3D perception.