About Me
I’m currently a Research Scientist Intern at Meta Superintelligence Labs, where I work on red-teaming personal agents. Before joining Meta, I completed my Ph.D. at the Graduate School of AI, KAIST, under the supervision of Prof. Eunho Yang at the Machine Learning and Intelligence Lab.
Research Interests
My research focuses on building reliable self-evolving AI that learns from its own experience and remains robust and safe in real-world environments.
- Self-Evolving Agents: Learning from own failures through RL (Preprint’26), Graph-structured skill memory for learning unknown actions (Preprint’26).
- Reliability: Red-teaming personal agents (Preprint’26), Robust visual reasoning (Preprint’26), Robust learning with biased data (ICML’23, ICML’24, NeurIPS’24), Jailbreaking (CVPR’25).
News
- Sep 2026
- Co-organizing AdvML-Frontiers × CoTMA at COLM 2026.
- Jun 2026
- Joined Meta Superintelligence Labs as a Research Scientist Intern (Menlo Park).
- Jun 2025
- One paper accepted to ICCV 2025.
- Jan 2025
- One paper accepted to ICLR 2025.
- Sep 2024
- One paper accepted to NeurIPS 2024.
- May 2024
- One paper accepted to ICML 2024.
- Apr 2023
- One paper accepted to ICML 2023.
Preprints
-
X-AgentRed: Red-Teaming Personal Agents in Evolving Multi-Service Ecosystems
Yeonsung Jung, Zhiyuan Wang, Zifan Wang, Vítor Albiero, Benjamin Newman, Edoardo Debenedetti, Ivan Evtimov, Ninareh Mehrabi -
From Action Discovery to Task Completion: Learning Unknown Actions and Reusable Skills for Self-Evolving Agents
Yeonsung Jung, Trilok Padhi, Eunho Yang, Ninareh Mehrabi -
Looks the Same, Answers Differently: Flip-Direction Steering for Robust Vision-Language Reasoning [paper]
Yeonsung Jung, Joonhyun Jeong, Hoang Pham, Joowon Kim, Yoonsik Park, Viet Dac Lai†, Eunho Yang† -
Co-Evolving Agents: Learning from Failures as Hard Negatives [paper]
Yeonsung Jung, Trilok Padhi, Sina Shaham, Dipika Khullar, Joonhyun Jeong, Ninareh Mehrabi†, Eunho Yang† -
Web Agents Are Still Greedy: Progress-Aware Action Generation and Selection via Meta-Plan
Joonhyun Jeong, Gilhyun Nam, Yeonsung Jung, Eunho Yang -
MeZO-A3dam: Memory-efficient Zeroth-order Adam with Adaptivity Adjustments for Fine-tuning LLMs
Sihwan Park*, Jihun Yun*, Sung-Yub Kim, June Yong Yang, Yeonsung Jung, Souvik Kundu, Kyungsu Kim, Eunho Yang -
3D Scene Decomposition Under Occlusion via Multi-View-Aware Inpainting
Heecheol Yun, Yeonsung Jung, Eunho Yang
Conference Publications
-
Early Timestep Zero-Shot Candidate Selection for Instruction-Guided Image Editing [paper]
Joowon Kim, Ziseok Lee, Donghyeon Cho, Sanghyun Jo, Yeonsung Jung, Kyungsu Kim, Eunho Yang
ICCV 2025 -
Preserve or Modify? Context-Aware Evaluation for Balancing Preservation and Modification in Text-Guided Image Editing [paper]
Yoonjeon Kim*, Soohyun Ryu*, Yeonsung Jung, Hyunkoo Lee, Joowon Kim, June Yong Yang, Jaeryong Hwang, Eunho Yang
CVPR 2025 -
Playing the Fool: Jailbreaking LLMs and Multimodal LLMs with Out-of-Distribution Strategy [paper]
Joonhyun Jeong, Seyun Bae, Yeonsung Jung, Jaeryong Hwang, Eunho Yang
CVPR 2025 -
LANTERN: Accelerating Visual Autoregressive Models with Relaxed Speculative Decoding [paper]
Doohyuk Jang*, Sihwan Park*, June Yong Yang, Yeonsung Jung, Jihun Yun, Souvik Kundu, Sung-Yub Kim†, Eunho Yang†
ICLR 2025 -
A Simple Remedy for Dataset Bias via Self-Influence: A Mislabeled Sample Perspective [paper]
Yeonsung Jung*, Jaeyun Song*, June Yong Yang, Jin-Hwa Kim, Sung-Yub Kim, Eunho Yang
NeurIPS 2024 -
PruNeRF: Segment-Centric Dataset Pruning via 3D Spatial Consistency [paper]
Yeonsung Jung, Heecheol Yun, Joonhyung Park, Jin-Hwa Kim†, Eunho Yang†
ICML 2024 -
Fighting Fire with Fire: Contrastive Debiasing without Bias-free Data via Generative Bias-transformation [paper]
Yeonsung Jung, Hajin Shim, June Yong Yang, Eunho Yang
ICML 2023 -
Scalable Anti-TrustRank with Qualified Site-level Seeds for Link-based Web Spam Detection [paper]
Joyce Jiyoung Whang, Yeonsung Jung, Seonggoo Kang, Dongho Yoo, Inderjit S. Dhillon
The Web Conf. Workshop on CyberSafety: Computational Methods in Online Misbehavior 2020 -
Fast Asynchronous Anti-TrustRank for Web Spam Detection [paper]
Joyce Jiyoung Whang, Yeonsung Jung, Inderjit S. Dhillon, Seonggoo Kang, Jungmin Lee
WSDM Workshop on MIS2: Misinformation and Misbehavior Mining on the Web 2018
Work Experience
- Research Scientist Intern, Meta Superintelligence Labs, Menlo Park (Jun 2026–Nov 2026)
- Agentic red-teaming for multi-service ecosystems.
-
Postdoctoral Researcher, Machine Learning and Intelligence Lab, KAIST, Daejeon, South Korea (Mar 2026–May 2026)
- External Collaborator, NAVER AI, Seongnam, South Korea (Sep 2023–Feb 2024)
- Robust learning for neural radiance fields; PruNeRF (ICML 2024).
- Research Intern, NAVER Search, Seongnam, South Korea (Jul 2019–Aug 2019)
- Graph-based ranking and search relevance for production systems.
Academic Service
- Workshop Organizer
- Conference Reviewer
- Neural Information Processing Systems (NeurIPS)
- International Conference on Machine Learning (ICML)
- International Conference on Learning Representations (ICLR)
- Computer Vision and Pattern Recognition (CVPR)
- International Conference on Computer Vision (ICCV)
- Association for the Advancement of Artificial Intelligence (AAAI)
- Artificial Intelligence and Statistics (AISTATS)
- NAACL Workshop on TrustNLP
Teaching Experience
SAMSUNG DS (2020–2022)
LG Electronics (2022)
HD Hyundai (2023)
LG Innotek (2022–2024)
LG Chem (2022)
DSME (2022–2023)