Yeda Song

Hi! I am a Master's student in the interdisciplinary AI program at Seoul National University, under the supervision of Gunhee Kim.

My research interest lies in deep reinforcement learning (RL), particularly in devising intelligent agents capable of solving novel tasks. I believe such a goal can be achieved by learning proper abstractions, utilizing language models, and taking model-based approaches.

Email  /  CV  /  Github

profile photo
MPChat: Towards Multimodal Persona-Grounded Conversation
Jaewoo Ahn, Yeda Song, Sangdoo Yun, Gunhee Kim,
ACL, 2023
CODE / arXiv

We construct a multimodal persona-grounded dialogue dataset, MPChat, accompanied with entailment labels. The multimodal persona consists of image-text pairs that represent one's episodic memories. We show the role of visual modality is crucial in MPChat through three benchmark tasks.

Ongoing Projects
Skill Discovery with Human Feedback: investigating the potential of human feedback for discovering complex temporal abstractions, or ‘skills’, within the context of RL
Offline Model-Based RL for Distributional Shift: tackling the problem of distributional shift in offline RL (batch RL) with model-based RL approaches
Seoul National University
Master's Student in Artificial Intelligence
Mar. 2022 - Present
Seoul National University
B.S. in Statistics
B.S. in Artificial Intelligence
Mar. 2017 - Feb. 2022
Hong Kong University of Science and Technology
Exchange Student
Sep. 2019 - Feb. 2020
Seoul Science High School Mar. 2014 - Feb. 2017
Anomaly Analysis Lab, Alchera Inc.
Machine Learning Researcher
Jun. 2021 - Aug. 2021
Multiscale Methods in Statistics Lab, Seoul National University
Research Intern
Mar. 2021 - Jun. 2021
Bioinformatics and Biostatistics Lab, Seoul National University
Research Intern
Jan. 2020 - Jun. 2020
Honors and Awards
Presidential Science Scholarship Mar. 2017 - Feb. 2021
Excellence Scholarship for Freshmen (Dept. of Statistics) Spring 2017
Hanseong Nobel Scholarship (Sector: Mathematics) Mar. 2015 - Feb. 2017

This website is built with Jon Barron's template.