Inseop Chung

Machine Learning Researcher

Building robust and intelligent learning systems for complex real-world environments.

I'm a Staff Researcher at the AI Center of Samsung Electronics, where I work on machine learning for semiconductor manufacturing, with a focus on advanced process control (APC), process modeling, and data-driven process optimization.

My research broadly explores robust, adaptive, and generalizable machine learning for complex and dynamic systems. During my Ph.D. at Seoul National University, I studied computer vision and generalizable machine learning, including domain adaptation, domain generalization, test-time adaptation, and knowledge distillation. I later extended this work to robotics and embodied AI, exploring how perception, language, and action can support intelligent autonomous systems.

I received my Ph.D. from the Graduate School of Convergence Science and Technology (GSCST), Department of Intelligence and Information at Seoul National University (SNU), under the supervision of Professor Nojun Kwak. I was a member of the Machine Intelligence and Pattern Analysis Lab (MIPAL), led by Professor Kwak.

Today, I apply and extend these machine learning foundations to semiconductor manufacturing, tackling process prediction, monitoring, and control. I am especially interested in learning from heterogeneous equipment sensor data and process measurements while remaining reliable under changing process conditions and limited data.

Ultimately, I aim to connect advances in machine learning with real-world applications and build intelligent systems that can understand, predict, and optimize complex physical processes.

Email / CV / Google Scholar / LinkedIn / GitHub

Research Interests

My work connects foundational machine learning with applications in complex physical systems. My current focus is semiconductor manufacturing, supported by a broader research foundation in generalization and adaptation, and experience in computer vision, robotics, and embodied AI.

Machine Learning for Semiconductor Manufacturing
Advanced process control (APC), process modeling, metrology prediction, process monitoring, and data-driven optimization.

Robust and Generalizable Machine Learning
Domain adaptation and generalization, test-time adaptation, transfer learning, knowledge distillation, and learning under distribution shifts.

Computer Vision, Robotics, and Embodied AI
Visual recognition, object detection, vision-language-action models, robot learning, and embodied intelligence.

Emerging Interests
Physics-informed machine learning, multimodal foundation models, world models, and AI for complex physical systems.

News

[2026.09] Our paper H-VLA was accepted to CoRL 2026 (Nov. 2026).
[2026.03] Released an open-source project, Apartment-Trade-Up-Calculator using Google Antigravity.
[2025.08] Our paper on Continual Test-Time Adaptation was published in IEEE Access.
[2024.12] Joined the AI Center at Samsung Electronics as a Staff Researcher.
[2024.06] One paper accepted to CVPR 2024 and another to the ICLR 2024 Workshop.
[2024.03] Joined the Samsung Advanced Institute of Technology (SAIT) as a Staff Researcher.
[2024.02] Received my Ph.D. from Seoul National University.

Publications

	KAPPA(κ): Keyframe-Anchored Predictive Policy for Vision-Language-Action Jiwon Lee, Inseop Chung*, Sujin Jang†, Jun Won Choi† Preprint* KAPPA predicts a latent representation of the next keyframe as a future visual milestone and conditions each action chunk on this prediction, anchoring continuous control to a sparse sequence of task-critical milestones. It achieves 98.4% overall success (96.9% on long-horizon tasks) on LIBERO and 57.3% on RoboCasa atomic tasks, with real-robot gains of +5.0% and +17.5% for in-distribution and out-of-distribution settings, respectively.
	H-VLA: Hierarchical Vision-Language-Action Model with Key-Action Reasoning and Motion Planning in a Unified Action Space Xiongfeng Peng, Lu Xu, Yandong Wang, Jiaqian Yu, Zirui Zheng, Yamin Mao, Weiming Li, Inseop Chung, Hyun-woong Cho, Jaewook Yoo, Dongwook Lee, Daehyun Ji, Chao Zhang Conference on Robot Learning (CoRL), 2026.Nov arXiv Decouples key-action reasoning from motion generation via a unified camera-centric action space, giving consistent representations across embodiments and viewpoints. Achieves 91% on SimplerEnv Google Robot and up to +52 points on Agilex real-robot OOD tasks.
	Mitigating the Bias in the Model for Continual Test-Time Adaptation Inseop Chung, Kyomin Hwang, Jayeon Yoo, Nojun Kwak IEEE Access, 2025.Aug arXiv This work improves Continual Test-Time Adaptation by introducing class-wise feature clustering and source-aligned prototype anchoring. It dynamically reduces prediction bias and enhances model robustness under distribution shifts.
	Open Domain Generalization with a Single Network by Regularization Exploiting Pretrained Features Inseop Chung, Kiyoon Yoo, Nojun Kwak ICLR 2024 Workshop, 2024.May arXiv We propose a single-network method for Open Domain Generalization that improves unseen domain generalization and unknown class detection. It leverages a linear-probed head and two regularization terms to refine the classifier while preserving pre-trained features, enabling efficient and robust generalization.
	What, How, and When Should Object Detectors Update in Continually Changing Test Domains? Jayeon Yoo, Dongkwan Lee, Inseop Chung, Donghyun Kim, Nojun Kwak CVPR, 2024.Jun arXiv We propose a lightweight, architecture-agnostic test-time adaptation method for object detection under continual domain shifts. By updating only small adaptor modules and introducing a class-wise feature alignment strategy, our approach achieves efficient, robust adaptation without modifying the backbone, improving mAP on standard benchmarks while preserving real-time performance.
	End-to-End Multi-Object Detection with a Regularized Mixture Model Jaeyoung Yoo, Hojun Lee, Seunghyeon Seo, Inseop Chung, Nojun Kwak ICML, 2023.Jul arXiv We introduce D-RMM, an end-to-end multi-object detector trained with negative log-likelihood and a novel regularization loss. By modeling detection as density estimation, our method removes training heuristics, improves confidence reliability, and outperforms prior end-to-end detectors on MS COCO.
	Exploiting Inter-pixel Correlations in Unsupervised Domain Adaptation for Semantic Segmentation Inseop Chung, Jayeon Yoo, Nojun Kwak WACV 2023 Workshop (Best Paper Award), 2023.Jan arXiv We enhance UDA for semantic segmentation by transferring inter-pixel correlations from the source to the target domain using a self-attention module. Trained only on the source, this module guides the segmentation network on the target domain beyond noisy pseudo labels, improving performance on standard benchmarks.
	X-MAS: Extremely Large-Scale Multi-Modal Sensor Dataset for Outdoor Surveillance in Real Environments DongKi Noh, Chang Ki Sung, Taeyoung Uhm, Wooju Lee, Hyungtae Lim, Jaeseok Choi, Kyuewang Lee, Dasol Hong, Daeho Um, Inseop Chung, Hochul Shin, Min-Jung Kim, Hyoung-Rock Kim, Seung-Min Baek, Hyun Myung IEEE Robotics and Automation Letters (RA-L), 2023.Jan We introduce X-MAS, a large-scale multi-modal dataset for outdoor surveillance with 500K+ annotated image pairs across five sensor types.
	Unsupervised Domain Adaptation for One-Stage Object Detector Using Offsets to the Bounding Box Jayeon Yoo, Inseop Chung, Nojun Kwak ECCV, 2022.Oct arXiv We propose OADA, an offset-aware domain adaptive object detector that conditions feature alignment on bounding box offsets to reduce negative transfer. This simple yet effective method improves both discriminability and transferability, achieving state-of-the-art results in domain adaptive detection.
	Dummy Prototypical Networks for Few-Shot Open-Set Keyword Spotting Byeonggeun Kim, Seunghan Yang, Inseop Chung, Simyung Chang INTERSPEECH, 2022.Sep arXiv We propose D-ProtoNets, a metric learning-based method for few-shot open-set keyword spotting. Using dummy prototypes, our approach effectively rejects unknown classes and outperforms recent FSOSR methods on both splitGSC and miniImageNet benchmarks.
	Personalized Keyword Spotting through Multi-task Learning Seunghan Yang, Byeonggeun Kim, Inseop Chung, Simyung Chang INTERSPEECH, 2022.Sep arXiv We propose PK-MTL, a multi-task learning framework for personalized keyword spotting that jointly learns keyword spotting and speaker verification.
	Maximizing Cosine Similarity Between Spatial Features for Unsupervised Domain Adaptation in Semantic Segmentation Inseop Chung, Daesik Kim, Nojun Kwak WACV, 2022.Jan arXiv We present a cosine similarity-based method for unsupervised domain adaptation in semantic segmentation. By aligning source and target features via a class-wise feature dictionary and similarity maximization, our approach reduces domain gaps and improves performance on standard UDA benchmarks.
	Multi-modal Object Detection, Tracking, and Action Classification for Unmanned Outdoor Surveillance Robots Kyuewang Lee, Inseop Chung*, Daeho Um, Jaeseok Choi, Yeji Song, Seunghyeon Seo, Nojun Kwak, Jin Young Choi ICCAS*, 2021.Oct We present a real-time multi-modal framework for outdoor robots that detects, tracks, and classifies human actions using RGB, thermal, and LiDAR data.
	Training Multi-Object Detector by Estimating Bounding Box Distribution for Input Image Jaeyoung Yoo, Hojun Lee, Inseop Chung*, Geonseok Seo, Nojun Kwak ICCV*, 2021.Oct arXiv We propose MDOD, a novel object detector that formulates multi-object detection as bounding box density estimation using a mixture model, simplifying training and improving performance.
	Feature Fusion for Online Mutual Knowledge Distillation Jangho Kim, Minsugn Hyun, Inseop Chung, Nojun Kwak ICPR, 2021.Jan arXiv We propose Feature Fusion Learning (FFL), a mutual knowledge distillation framework that combines features from diverse sub-networks to train a stronger fused classifier, improving both overall and individual network performance.
	Feature-map-level Online Adversarial Knowledge Distillation Inseop Chung, SeongUk Park, Jangho Kim, Nojun Kwak ICML, 2020.Jul Proceedings We propose an online knowledge distillation method that aligns feature maps via adversarial training, enabling networks to learn from each other's internal representations and improving performance, especially in mixed-size network setups.

Education
	Ph.D. in Engineering, Seoul National University, 2019–2024 Graduate School of Convergence Science and Technology, Department of Intelligence and Information Advisor: Nojun Kwak
	B.S. in Creative Technology Management & Computer Science, Yonsei University, 2012–2019 Exchange Student Program at Technical University of Munich (2016–2017)

Work Experience
	Staff Researcher, AI Center, Samsung Electronics, 2024.Dec–Present Researching machine learning for semiconductor manufacturing, with a focus on advanced process control (APC), process modeling, metrology prediction, and data-driven process optimization. My work explores robust learning from equipment sensor data and process measurements under changing process conditions. Staff Researcher, Samsung Advanced Institute of Technology (SAIT), Samsung Electronics, 2024.Mar–2024.Dec Expanded my research into robotics and embodied AI, including vision-language-action models that integrate perception, language, and action for autonomous systems.
	Research Intern, Qualcomm AI Research Korea, 2022.Jan–2022.Jul Developed few-shot learning methods for keyword spotting to recognize new keywords from minimal examples, enhancing speech recognition performance.
	Research Intern, Naver Webtoons Corp., 2020.Jun–2020.Dec Researched unsupervised domain adaptation for semantic segmentation, improving model generalization across diverse webtoon styles through novel adaptation strategies.

Patents

Domestic Patents

Device for Unsupervised Domain Adaptation in Semantic Segmentation Exploiting Inter-pixel Correlations and Driving Method Thereof

Device for Regression Scale-aware Cross-domain Object Detection and Driving Method Thereof

International Patents

Method and Apparatus with Online Task Planning

Method and Apparatus with Micro-Action Determination

Multi-Task Learning for Personalized Keyword Spotting

Dummy Prototypical Networks For Few-Shot Open-Set Keyword Spotting

Funding

Samsung Electronics Ph.D. Student Sponsorship Program (2021-2024)

Honour & Awards

The best paper award in the 3rd Workshop on Real-World Surveillance: Applications and Challenges at WACV2023, $1,000 (2023)

BK21 Graduate School Innovation Project Colloquium Outstanding Graduate Student Award, ￦500,000 (2021)

Qualcomm Innovation Fellowship Korea 2020 Final Session, Finalist (2020)

Samsung Electronics - Seoul National University, Industry-Academic Cooperation Research Award, ￦1,500,000 Won (2020)

Special Award, Yonsei Programming Contest 2018 (Spring Semester 2018)

Honours, tuition support scholarship, Yonsei University (Spring Semester 2016)

Website template courtesy of Jon Barron