-
16
- Application-aware System Optimization Lab (Professor: Park, Yongjun (박영준)) Publishes Two Papers at CGO and Chairs a Sess
- Application-aware System Optimization Lab (Professor: Park, Yongjun (박영준)) Publishes Two Papers at CGO and Chairs a Session Students from the Application-aware System Optimization Lab (Professor: Park, Yongjun (박영준)) presented two papers at the International Symposium on Code Generation and Optimization 2025 (CGO ‘25), one of the top international conferences in the field of compilers. Additionally, Prof. Park, Yongjun (박영준) served as the session chair for the Architectures and Code Generation session at the conference. The first paper, titled “CUrator: An Efficient LLM Execution Engine with Optimized Integration of CUDA Libraries,” proposes a technique for efficiently executing large language model (LLM) inference by leveraging cuBLAS and CUTLASS libraries on various modern GPUs. This research demonstrates peak inference performance for LLMs across multiple GPUs and is expected to guide the future direction of next-generation optimization frameworks. Paper link The second paper, titled “Accelerating LLMs using an Efficient GEMM Library and Target-Aware Optimizations on Real-world PIM Devices,” introduces an optimized GEMM library designed for Processing-in-Memory (PIM) architectures and proposes additional optimization techniques to accelerate LLM inference. This study effectively utilizes PIM architectures to address the inference slowdown caused by the high data requirements of large language models, playing a crucial role in overcoming this challenge. Paper link
- 첨단컴퓨팅학부 2025.03.25
-
15
- Yonsei University’s School of Computing Professors Have Papers Accepted at CHI 2025, the World’s Leading HCI Conference
- Yonsei University’s School of Computing Professors Have Papers Accepted at CHI 2025, the World’s Leading HCI Conference Professors from Yonsei University’s School of Computing have had their research papers accepted at CHI 2025, the ACM Conference on Human Factors in Computing Systems, the most prestigious academic conference in the field of Human-Computer Interaction (HCI). CHI 2025 is a premier international conference where pioneering research in HCI is presented. It will take place in Yokohama, Japan, from April 26 to May 1, 2025. The theme of CHI 2025 is "Ikigai", meaning "a purpose in life." As research on how technology can positively impact human life is increasingly important, the contributions of Yonsei University researchers stand out even more. Their papers address research topics that have a tangible impact on people’s lives, making their acceptance at CHI 2025 a particularly meaningful achievement. This accomplishment further solidifies Yonsei University’s global leadership in HCI and AI research. The university remains committed to advancing technologies that bring real benefits to people’s lives through ongoing research and innovation. List of accepted papers: 1. Crafting Champions: An Observation Study of Esports Coaching Processes - Hanbyeol Lee*, Erica Kleinman*, Namsub Kim, Sangbeom Park, Casper Harteveld, Byungjoo Lee 2. Data Formulator 2: Iterative Creation of Data Visualizations, with AI Transforming Data Along the Way - Chenglong Wang, Bongshin Lee, Steven M. Drucker, Dan Marshall, Jianfeng Gao 3. DataSentry: Building Missing Data Management System for In-the-Wild Mobile Sensor Data Collection through Multi-Year Iterative Design Approach - Yugyeong Jung, Hei Yiu Law, Hazel Hadong Lee, Junmo Lee, Bongshin Lee, Uichin Lee 4. FluidTrack: Investigating Child-Parent Collaborative Tracking for Pediatric Voiding Dysfunction Management - Junhyung Moon, Sukhyun Lee, Youngchan Kim, Juhee Go, Han Mo Ku, Yeohyun Jung, Seonyeong Hwang, Bongshin Lee, Yong Seung Lee, Hyun-Kyung Lee, Kyoungwoo Lee*, Eun Kyoung Choe* 5. Hardware-Embedded Pointing Transfer Function Capable of Canceling OS Gains - Seonho Kim, Munjeong Kim, Jonghyun Kim, Donghyeon Kang, Sunjun Kim, Byungjoo Lee 6. Modeling User Performance in Multi-Lane Moving-Target Acquisition - Jonghyun Kim, Joongseok Kim, June-Seop Yoon, Hee-Seung Moon, Sunjun Kim, Byungjoo Lee 7. PlanTogether: Facilitating AI Application Planning Using Information Graphs and Large Language Models - Dae Hyun Kim*, Daeheon Jeong*, Shakhnozakhon Yadgarova, Hyungyu Shin, Jinho Son, Hariharan Subramonyam, Juho Kim In addition to the full papers mentioned above, two Late Breaking Work papers will also be presented as posters. [Late-Breaking Work (Poster)] 1. LLM Adoption in Data Curation Workflows: Industry Practices and Insights - Crystal Qian, Michael Xieyang Liu, Emily Reif, Grady Simon, Nada Hussein, Nathan Clement, James Wexler, Carrie J. Cai, Michael Terry, Minsuk Kahng 2. Who Helps the Helpers?: Complications and Considerations for ICT Instructors Teaching Older Adults - Jiwon Song, Bongshin Lee, Jinwook Seo*, Eun Kyoung Choe* #YonseiUniversity #CHI2025 #HCI #ResearchAchievement #AI #Ikigai Image Source:CHI 2025 Facebook (https://www.facebook.com/acmchi)
- 첨단컴퓨팅학부 2025.03.20
-
14
- Yonsei University’s School of Computing Has 9 Papers Accepted at CVPR 2025
- Yonsei University’s School of Computing Has 9 Papers Accepted at CVPR 2025 Professors from Yonsei University’s School of Computing have had 9 papers accepted at the Conference on Computer Vision and Pattern Recognition (CVPR) 2025. These papers present groundbreaking advancements in computer vision and pattern recognition, earning high praise from researchers worldwide. In particular, they introduce innovative methodologies applicable to cutting-edge fields such as deep learning-based image understanding, autonomous driving, and medical image analysis, drawing significant attention from both academia and industry. CVPR is one of the most prestigious international conferences, where top AI and computer vision researchers gather annually to present and discuss state-of-the-art research. With its highly competitive acceptance rate, CVPR stands as a premier venue for AI research. The acceptance of these papers once again demonstrates the global research competitiveness of Yonsei University’s School of Computing and solidifies its position as a leading institution in AI research. List of accepted papers: 1. Distilling Spectral Graph for Object-Context Aware Open-Vocabulary Semantic Segmentation - Chanyoung Kim, Dayun Ju, Woojung Han, Ming-Hsuan Yang, Seong Jae Hwang 2. EditSplat: Multi-View Fusion and Attention-Guided Optimization for View-Consistent 3D Scene Editing with 3D Gaussian Splatting - Dong In Lee, Hyeongcheol Park, Jiyoung Seo, Eunbyung Park, Hyunje Park, Ha Dam Baek, Shin Sangheon, Sangmin Kim, Sangpil Kim 3. Generative Densification: Learning to Densify Gaussians for High-Fidelity Generalizable 3D Reconstruction - Seungtae Nam*, Xiangyu Sun*, Gyeongjin Kang, Younggeun Lee, Seungjun Oh, Eunbyung Park 4. Latent space Super-Resolution for Higher-Resolution Image Generation with Diffusion Models - Jinho Jeong, Sangmin Han, Jinwoo Kim, Seon Joo Kim 5. Omni-RGPT: Unifying Image and Video Region-level Understanding via Token Marks - Miran Heo, Min-Hung Chen, De-An Huang, Sifei Liu, Subhashree Radhakrishnan, Seon Joo Kim, Yu-Chiang Frank Wang, Ryo Hachiuma - https://miranheo.github.io/omni-rgpt/ 6. ORIDa: Object-centric Real-world Image Composition Dataset - Jinwoo Kim, Sangmin Han, Jinho Jeong, Jiwoo Choi, Dongyoung Kim, Seon Joo Kim 7. SelfSplat: Pose-Free and 3D Prior-Free Generalizable 3D Gaussian Splatting - Gyeongjin Kang*, Jisang Yoo*, Jihyeon Park, Seungtae Nam, Hyeonsoo Im, Sangheon Shin, Sangpil Kim, Eunbyung Park 8. Spatial Transport Optimization by Repositioning Attention Map for Training-Free Text-to-Image Synthesis - Woojung Han, Yeonkyung Lee, Chanyoung Kim, Kwanghyun Park, Seong Jae Hwang 9. Your Large Vision-Language Model Only Needs A Few Attention Heads for Visual Grounding - Seil Kang, Jinyoung Kim, Junhyeok Kim, Seong Jae Hwang Image Source:CVPR Official Website (https://cvpr.thecvf.com)
- 첨단컴퓨팅학부 2025.03.20
-
13
- Yonsei University’s School of Computing Has 12 Papers Accepted at ICLR 2025
- Yonsei University’s School of Computing Has 12 Papers Accepted at ICLR 2025 Professors from Yonsei University’s School of Computing have had 12 papers accepted at the International Conference on Learning Representations (ICLR) 2025. ICLR is a top-tier international conference that focuses on cutting-edge research in deep learning and representation learning. Every year, AI experts from around the world gather to share groundbreaking research findings. With this achievement, Yonsei University’s School of Computing has once again demonstrated its global research competitiveness and contributed to elevating the status of Korean AI research on the international stage. List of accepted papers: 1. Adaptive Energy Alignment for Accelerating Test-Time Adaptation - Wonjeong Choi, Do-Yeon Kim, Jungwuk Park, Jungmoon Lee, Younghyun Park, Dong-Jun Han, and Jaekyun Moon 2. Anti-Exposure Bias in Diffusion Models via Prompt Learning (Spotlight Presentation) - Junyu Zhang, Daochang Liu, Eunbyung Park, Shichao Zhang, Chang Xu 3. Asynchronous Federated Reinforcement Learning with Policy Gradient Updates: Algorithm Design and Convergence Analysis - Guangchen Lan, Dong-Jun Han, Abolfazl Hashemi, Vaneet Aggarwal, Christopher G. Brinton 4. Decentralized Sporadic Federated Learning: A Unified Algorithmic Framework with Convergence Guarantees (Spotlight Presentation) - Shahryar Zehtabi, Dong-Jun Han, Rohit Parasnis, Seyyedali Hosseinalipour, Christopher G. Brinton 5. Model-based Offline Reinforcement Learning with Lower Expectile Q-Learning - Kwanyoung Park, Youngwoon Lee 6. PIG: Physics-Informed Gaussians as Adaptive Parametric Mesh Representations - Namgyu Kang*, Jaemin Oh*, Youngjoon Hong, Eunbyung Park 7. PRISM: Privacy-Preserving Improved Stochastic Masking for Federated Generative Models - Kyeongkook Seo, Dong-Jun Han*, Jaejun Yoo* 8. See What You Are Told: Visual Attention Sink in Large Multimodal Models - Seil Kang*, Jinyeong Kim*, Junhyeok Kim, Seong Jae Hwang 9. SEMDICE: Off-policy State Entropy Maximization via Stationary Distribution Correction Estimation - Jongmin Lee*, Meiqi Sun*, Pieter Abbeel 10. Spread Preference Annotation: Direct Preference Judgment for Efficient LLM Alignment (Oral Presentation (207/11672=1.77%)) - Dongyoung Kim, Kimin Lee, Jinwoo Shin, and Jaehyung Kim 11. Unlocking the Potential of Model Calibration in Federated Learning - Yun-Wei Chu, Dong-Jun Han, Seyyedali Hosseinalipour, Christopher G. Brinton 12. Web Agents with World Models: Learning and Leveraging Environment Dynamics in Web Navigation - Hyungjoo Chae, Namyoung Kim, Kai Tzu-iunn Ong, Minju Gwak, Gwanwoo Song, Jihoon Kim, Sunghwan Kim, Dongha Lee, Jinyoung Yeo Image Source: ICLR Official Website (https://iclr.cc)
- 첨단컴퓨팅학부 2025.03.20
-
12
- Dongjin Kang andSunghwan Kim from the Department of Artificial Intelligence DLI Lab have won the ACL 2024 (2024-08-28)
- Dongjin Kang andSunghwan Kim from the Department of Artificial Intelligence DLI Lab have won the ACL 2024 Outstanding Paper Award. “Can Large Language Models be Good Emotional Supporter? Mitigating Preference Bias on Emotional Support Conversation”, {Dongjin Kang, Sunghwan Kim}, Taeyoon Kwon, Seungjun Moon, Hyunsouk Cho, Youngjae Yu, Dongha Lee, Jinyoung Yeo, ACL 2024, Outstanding Paper Award Link: : https://arxiv.org/abs/2402.13211
- 첨단컴퓨팅학부 2025.03.20
-
11
- Professor Jinkyu Jeong's lab (Scalable Systems Software Lab) presented two papers at top-tier international (2024-08-02)
- Professor Jinkyu Jeong's lab (Scalable Systems Software Lab) presented two papers at top-tier international conferences in the field of systems software (OSDI '24, USENIX ATC '24) In July 2024, the Scalable Systems Software Lab at Yonsei University, led by Professor Jeong, presented a paper at the 18th USENIX Symposium on Operating Systems Design and Implementation (OSDI '24), the top international conference in the field of operating systems. The paper titled “Identifying On-/Off-CPU Bottlenecks Together with Blocked Samples” introduces the Blocked Samples technique, a key innovation that greatly simplifies application performance profiling in today’s increasingly diverse and complex computer systems. The research presents two application performance profilers, bperf and BCOZ, which leverage this technique. These profilers are valuable tools for optimization as they pinpoint bottlenecks that lead to performance improvements when addressed. The research team demonstrated the utility of these profilers by profiling and optimizing the performance of large-scale language model inference and NoSQL big data storage systems. Additionally, the Blocked Samples technique reduces application performance interference by up to 17 times compared to similar existing profiler tools. Additionally, Professor Jeong’s research team presented a paper at the 2024 USENIX Annual Technical Conference (USENIX ATC '24), a flagship international conference in the field of system software, which was held alongside OSDI '24. The paper titled “A Secure, Fast, and Resource-Efficient Serverless Platform with Function REWIND” identifies security issues caused by container (or sandbox) reuse techniques used to enhance performance in commercial serverless cloud platforms like Amazon Lambda and Google Cloud Functions. The paper introduces the REWIND technique, which addresses these security concerns while simultaneously improving performance and reducing memory usage. This approach selectively rewinds only the memory and file regions that could cause security issues after executing a serverless function within a serverless container. By doing so, it eliminates any residual user privacy data, ensuring security, while significantly reducing the memory usage required to maintain this security. The research team demonstrated that, across various real-world cloud workloads, the proposed technique maintains near-zero performance loss compared to less secure execution methods and reduces memory usage by more than half. Links: Identifying On-/Off-CPU Bottlenecks Together with Blocked Samples, https://www.usenix.org/conference/osdi24/presentation/ahn A Secure, Fast, and Resource-Efficient Serverless Platform with Function REWIND, https://www.usenix.org/conference/atc24/presentation/song
- 첨단컴퓨팅학부 2025.03.20
-
10
- Professor Seong Jae Hwang's Research Team Selected as Highlight Paper at CVPR 2024 (2024-05-23)
- A paper by Professor Seong Jae Hwang's research team, titled "EAGLE: Eigen Aggregation Learning for Object-Centric Unsupervised Semantic Segmentation," has been selected as a Highlight Paper at CVPR 2024, the top-tier international conference in the field of AI and computer vision. This achievement places it in the top 2.8% of all submitted papers. This research addresses the issue that existing methods do not consider object-level representation during training, proposing a new methodology to implement unsupervised semantic segmentation with object-centric representation. This method uses the eigenbasis of the Graph Laplacian to obtain clues about objects and conducts contrastive learning based on these clues. The research was conducted under the guidance of Professor Seong Jae Hwang(Dept. of AI), with contributions from Chanyoung Kim (Dept. of AI), Woojung Han (Dept. of CS), and Dayun Ju (Dept. of CS). Link: EAGLE: Eigen Aggregation Learning for Object-Centric Unsupervised Semantic Segmentation: https://arxiv.org/abs/2403.01482
- 첨단컴퓨팅학부 2025.03.20
-
9
- Two of Four Papers Presented by Yonsei Esports Lab at CHI 2024 Win Best Paper Honorable Mention Award (2024-05-03)
- Two of Four Papers Presented by Yonsei Esports Lab at CHI 2024 Win Best Paper Honorable Mention Award In May 2024, Yonsei Esports Lab, supervised by Professor Byungjoo Lee, will present four papers at CHI 2024, a premier international conference of Human-Computer Interaction. "Quantifying Wrist-Aiming Habits with A Dual-Sensor Mouse: Implications for Player Performance and Workload" paper presents a technique to quantify the extent of a player's wrist-aiming habits using a mouse equipped with two optical sensors and examines the relationship between wrist-aiming habits and player performance or workload. "Characterizing and Quantifying Expert Input Behavior in League of Legends" paper demonstrates a holistic pipeline of input behavior analysis from characterizing and quantifying the quality of League of Legends players’ input skills to providing actionable lessons with players based on visualization of input behavior. "Real-time 3D Target Inference via Biomechanical Simulation" paper proposes a novel approach that leverages biomechanical simulation to produce synthetic motion data, capturing a variety of movement-related factors, such as limb configurations and motor noise. "User Performance in Consecutive Temporal Pointing: An Exploratory Study" paper broadly explores user performance in a variety of Consecutive temporal pointing (CTP) scenarios and finds CTP is a unique task that cannot be considered as two ordinary temporal pointing processes. Among them, two papers won the Best Paper Honorable Mention Award, which is only awarded to the top 5% of papers. Link: Quantifying Wrist-Aiming Habits with A Dual-Sensor Mouse: Implications for Player Performance and Workload (Best Paper Honorable Mention Award) : https://programs.sigchi.org/chi/2024/program/content/146862 Characterizing and Quantifying Expert Input Behavior in League of Legends : https://programs.sigchi.org/chi/2024/program/content/148156 Real-time 3D Target Inference via Biomechanical Simulation (Best Paper Honorable Mention Award) : https://programs.sigchi.org/chi/2024/program/content/147400 User Performance in Consecutive Temporal Pointing: An Exploratory Study : https://programs.sigchi.org/chi/2024/program/content/147090
- 첨단컴퓨팅학부 2025.03.20
-
8
- DLI lab student Hyungjoo Chae from the Department of Artificial Intelligence publishes research results ... (2024-04-09)
- DLI lab student Hyungjoo Chae from the Department of Artificial Intelligence publishes research results on Large Language Models that surpasses Google DeepMind In April 2024, Hyungjoo Chae, a DLI lab student supervised by Professor Jinyoung Yeo, pre-published his latest research on Arxiv, drawing significant attention for its reasoning capabilities that far exceed those of Google DeepMind's "Self-Discover" framework. Notably, this research was introduced by famous Twitter influencers in the AI field, attracting intense interest with approximately 35,000 views. The research focuses on the ability of large language models (LLMs) to develop and utilize algorithms to solve problems, proposing a new methodology. The joint research team of Professor Jinyoung Yeo and Professor Youngjae Yu proved that this allows language models to solve complex problems more effectively. Moreover, the proposed methodology indicates a significant advantage by demonstrating cost-effective reasoning where massive and small language models work together. Link: https://huggingface.co/papers/2404.02575 https://twitter.com/_akhaliq/status/1775743181885186214
- 첨단컴퓨팅학부 2025.03.20
-
7
- Professor Baek Jong-duk's research team has been selected for the Korea Research Foundation's STEAM ... (2023-05-22)
- Professor Baek Jong-duk's research team has been selected for the Korea Research Foundation's STEAM Research Project (Bridge Convergence Research and Development) Professor Baek Jong-duk's research team has been finally selected for the Korea Research Foundation's STEAM Research Project. From April 2023 to December 2026 (3 years and 9 months), they will receive a total research grant of 3.025 billion Korean Won and lead the research team as the head of the 'High-Precision Robotic Surgery Image Guidance Technology AI Convergence Research Group.' In this research, the robotic surgery company, GoYoung Technology, the neurosurgery team at Seoul National University Hospital led by Professor Pi Ji-hoon, and the AI Graduate School at KAIST led by Professor Shim Hyun-jeong will participate as collaborative research institutions. Through this project, the research team will receive comprehensive support from technology development to commercialization, and in particular, during Phase 2 (January 2025 to December 2026), Professor Baek Jong-duk's academic startup, Barnes Imaging Co., Ltd., will participate in technology enhancement and joint development for business implementation
- 첨단컴퓨팅학부 2025.03.20