Kaist

Opening the Door to B Cell-Based Cancer-Rememberin..

  < (From left) KAIST Professor Jung Kyoon Choi, Dr. Jeong Yeon Kim, and Dr. Jin Hyeon An >   Neoantigens are unique markers that distinguish only cancer cells. By adding B cell reactivity, cancer vaccines can move beyond one-time attacks and short-term memory to become a long-term immunity that "remembers" cancer, effectively preventing recurrence. KAIST’s research team has developed an AI-based personalized cancer vaccine design technology that makes this possible and optimizes anticancer effects for each individual. KAIST announced on January 2nd that Professor Jung Kyoon Choi’s research team from the Department of Bio and Brain Engineering, in collaboration with Neogen Logic Co., Ltd., has developed a new AI model to predict neoantigens—a core element of personalized cancer vaccine development—and clarified the importance of B cells in cancer immunotherapy. The research team overcame the limitations of existing neoantigen discovery, which relied primarily on predicting T cell reactivity, and developed an AI-based neoantigen prediction technology that integrally considers both T cell and B cell reactivity. This technology has been validated through large-scale cancer genome data, animal experiments, and clinical trial data for cancer vaccines. It is evaluated as the first AI technology capable of quantitatively predicting B cell reactivity to neoantigens. Neoantigens are antigens composed of protein fragments derived from cancer cell mutations. Because they possess cancer-cell specificity, they have gained attention as a core target for next-generation cancer vaccines. Companies like Moderna and BioNTech developed COVID-19 vaccines using the mRNA platforms they secured while advancing neoantigen-based cancer vaccine technology, and they are currently actively conducting clinical trials for cancer vaccines alongside global pharmaceutical companies. However, current cancer vaccine technology is mostly focused on T cell-centered immune responses, presenting a limitation in that it does not sufficiently reflect the immune responses mediated by B cells. In fact, the research team of Professors Mark Yarchoan and Elizabeth Jaffee at Johns Hopkins University pointed out in Nature Reviews Cancer in May 2025 that “despite accumulating evidence regarding the role of B cells in tumor immunity, most cancer vaccine clinical trials still focus only on T cell responses.” The research team’s new AI model overcomes existing limitations by learning the structural binding characteristics between mutant proteins and B cell receptors (BCR) to predict B cell reactivity. In particular, an analysis of cancer vaccine clinical trial data confirmed that integrating B cell responses can significantly enhance anti-tumor immune effects in actual clinical settings. < Schematic Background of the Technology > Professor Jung Kyoon Choi stated, “Together with Neogen Logic Co., Ltd., which is currently commercializing neoantigen AI technology, we are conducting pre-clinical development of a personalized cancer vaccine platform and are preparing to submit an FDA IND＊ with the goal of entering clinical trials in 2027.” He added, “We will enhance the scientific completeness of cancer vaccine development based on our proprietary AI technology and push forward the transition to the clinical stage step-by-step.” ＊FDA IND: The procedure for obtaining permission from the U.S. Food and Drug Administration (FDA) to conduct clinical trials before administering a new drug to humans for the first time. Dr. Jeong Yeon Kim and Dr. Jin Hyeon An participated as co-first authors in this study. The research results were published in the international scientific journal Science Advances on December 3rd. ※ Paper Title: B cell–reactive neoantigens boost antitumor immunity, DOI: 10.1126/sciadv.adx8303    

Presenting a Brain-Like Next-Generation AI Semicon..

  < (From left) Professor Sanghun Jeon, Ph.D candidate Seungyeob Kim, Postdoctoral researcher Hongrae Cho, Ph.D candidates Sang-ho Lee and Taeseung Jung, and M.S candidate Seonjae Park >     With the advancement of Artificial Intelligence (AI), the importance of ultra-low-power semiconductor technology that integrates sensing, computation, and memory into a single unit is growing. However, conventional structures face challenges such as power loss due to data movement, latency, and limitations in memory reliability. A Korean research team has drawn international academic attention by presenting core technologies for an integrated ‘Sensor–Compute–Store’ AI semiconductor to solve these issues. KAIST announced on December 31st that Professor Sanghun Jeon’s research team from the School of Electrical Engineering presented a total of six papers at the ‘International Electron Devices Meeting (IEEE IEDM 2025)’—the world’s most prestigious semiconductor conference—held in San Francisco from December 8 to 10. Among these, the papers were simultaneously selected as a Highlight Paper and a Top Ranked Student Paper. Highlight Paper: Monolithically Integrated Photodiode–Spiking Circuit for Neuromorphic Vision with In-Sensor Feature Extraction [Link: https://iedm25.mapyourshow.com/8_0/sessions/session-details.cfm?scheduleid=255] Top Ranked Student Paper: A Highly Reliable Ferroelectric NAND Cell with Ultra-thin IGZO Charge Trap Layer; Trap Profile Engineering for Endurance and Retention Improvement [Link: https://iedm25.mapyourshow.com/8_0/sessions/session-details.cfm?scheduleid=124] The research on the M3D integrated neuromorphic vision sensor, selected as a highlight paper, is a semiconductor that stacks the human eye and brain within a single chip. Simply put, the sensors that detect light and the circuits that process signals like a brain are made into very thin layers and stacked vertically in one chip, implementing a structure where the process of 'seeing' and 'judging' occurs simultaneously. Through this, the research team completed the world's first "In-Sensor Spiking Convolution" platform, where AI computation technology that "sees and judges at the same time" takes place directly within the camera sensor. < Figure 1. Summary of research on vertically stacked optical signal-to-spike frequency converter for AI > < Figure 2. Representative diagram of the development of a 2T-2C near-pixel analog computing cell based on oxide thin-film transistors > Previously, this technology required several stages: capturing an image (sensor), converting it to digital (ADC), storing it in memory (DRAM), and then calculating (CNN). However, this new technology eliminates unnecessary data movement as the calculation happens immediately within the sensor. As a result, it has become possible to implement real-time, ultra-low-power Edge AI with significantly reduced power consumption and dramatically improved response speeds. Based on this approach, the research team presented six core technologies at the conference covering all layers of AI semiconductors, from input to storage. They simultaneously created neuromorphic semiconductors that operate like the brain using much less electricity while utilizing existing semiconductor processes, along with next-generation memory optimized for AI. First, on the sensor side, they designed the system so that judgment occurs at the sensor stage rather than having separate components for capturing images and calculating. Consequently, power consumption decreased and response speeds increased compared to the conventional method of taking a photo and sending it to another chip for calculation. < Figure 3. Schematic diagram of a next-generation biomimetic tactile system using neuromorphic devices > < Figure 4. Representative diagram of NC-NAND development research based on Ultra-thin-Mo and Sub-3.5 nm HZO > Furthermore, in the field of memory, they implemented a next-generation NAND flash that uses the same materials but operates at lower voltages, lasts longer, and can store data stably even when the power is turned off. Through this, they presented a foundational technology that satisfies the requirements for high-capacity, high-reliability, and low-power memory necessary for AI.   < Figure 5. Representative diagram of next-generation 3D FeNAND memory development research > < Figure 6. Representative diagram of research on charge behavior characterization and quantitative analysis methodology for next-generation FeNAND memory > Professor Sanghun Jeon, who led the research, stated, "This research is significant in that it demonstrates that the entire hierarchy can be integrated into a single material and process system, moving away from the existing AI semiconductor structure where sensing, computation, and storage were designed separately." He added, "Moving forward, we plan to expand this into a next-generation AI semiconductor platform that encompasses everything from ultra-low-power Edge AI to large-scale AI memory." Meanwhile, this research was conducted with support from basic research projects of the Ministry of Science and ICT and the National Research Foundation of Korea, as well as the Center for Heterogeneous Integration of Extreme-scale & Property Semiconductors (CH³IPS). It was carried out in collaboration with Samsung Electronics, Kyungpook National University, and Hanyang University.      

KAIST Awakens dormant immune cells inside tumors t..

  <(From Left) Professor Ji-Ho Park, Dr. Jun-Hee Han from the Department of Bio and Brain Engineering>   Within tumors in the human body, there are immune cells (macrophages) capable of fighting cancer, but they have been unable to perform their roles properly due to suppression by the tumor. KAIST researchers have overcome this limitation by developing a new therapeutic approach that directly converts immune cells inside tumors into anticancer cell therapies. KAIST (President Kwang Hyung Lee) announced on the 30th that a research team led by Professor Ji-Ho Park of the Department of Bio and Brain Engineering has developed a therapy in which, when a drug is injected directly into a tumor, macrophages already present in the body absorb it, produce CAR (a cancer-recognizing device) proteins on their own, and are converted into anticancer immune cells known as “CAR-macrophages.” Solid tumors—such as gastric, lung, and liver cancers—grow as dense masses, making it difficult for immune cells to infiltrate tumors or maintain their function. As a result, the effectiveness of existing immune cell therapies has been limited. CAR-macrophages, which have recently attracted attention as a next-generation immunotherapy, have the advantage of directly engulfing cancer cells while simultaneously activating surrounding immune cells to amplify anticancer responses. However, conventional CAR-macrophage therapies require immune cells to be extracted from a patient’s blood, followed by cell culture and genetic modification. This process is time-consuming, costly, and has limited feasibility for real-world patient applications. To address this challenge, the research team focused on “tumor-associated macrophages” that are already accumulated around tumors. They developed a strategy to directly reprogram immune cells in the body by loading lipid nanoparticles—designed to be readily absorbed by macrophages—with both mRNA encoding cancer-recognition information and an immunostimulant that activates immune responses. In other words, in this study, CAR-macrophages were created by “directly converting the body’s own macrophages into anticancer cell therapies inside the body.” <Figure . Schematic illustration of the strategy for in vivo CAR-macrophage generation and cancer cell eradication via co-delivery of CAR mRNA and immunostimulants using lipid nanoparticles (LNPs)>

Hemostasis in 1 Second... Boosting Survival Rates ..

  < (From top left) Professor Steve Park, Professor Sangyong Jon, (From bottom left) President Kwang-Hyung Lee, Ph.D canddiate Youngju Son, Ph.D candidate Kyusoon Park >   The leading cause of death due to injuries in war is excessive bleeding. A KAIST research team, in which an Army Major participated, has tackled this issue head-on. By developing a next-generation powder-type hemostatic agent that stops bleeding in one second just by spraying it, they have presented an innovative technology that will change the paradigm of combatant survivability. KAIST announced on December 29th that a joint research team led by Professor Steve Park from the Department of Materials Science and Engineering and Professor Sangyong Jon from the Department of Biological Sciences has developed a powder-type hemostatic agent that forms a powerful hydrogel barrier within approximately one second when sprayed on a wound. This technology reached a high level of perfection as a practical technology considering real combat environments, with an Army Major researcher directly participating in the study. By implementing characteristics that allow instant hardening even under extreme conditions such as combat and disaster sites due to high usability and storage stability, immediate emergency treatment is possible. Until now, patch-type hemostatic agents widely used in medical fields have had limitations in application to deep and complex wounds due to their flat structure, and were sensitive to temperature and humidity, posing limits on storage and operation. Accordingly, the research team developed a next-generation hemostatic agent in powder form that can be freely applied even to deep, large, and irregular wounds. They have secured versatility to respond to various types of wounds with a single powder. < AGCL powder development strategy and fabrication schematin/ Gelation speed and blood absorption capacity of AGCL powder > Existing powder hemostatic agents had limits in hemostatic capability as they functioned by physically absorbing blood to form a barrier. To solve this problem, the research team focused on the ionic reactions within the blood. The ‘AGCL powder’ developed this time has a structure that combines biocompatible natural materials such as Alginate and Gellan Gum (which react with calcium for ultra-fast gelation and physical sealing) and Chitosan (which bonds with blood components to enhance chemical and biological hemostasis). It reacts with cations such as calcium in the blood to turn into a gel state in one second, instantly sealing the wound. Furthermore, by forming a three-dimensional structure inside the powder, it can absorb blood amounting to more than 7 times its own weight (725％). Due to this, it quickly blocks blood flow even in high-pressure and excessive bleeding situations, and showed superior sealing performance compared to commercial hemostatic agents with a high adhesive strength of over '40kPa', a level of pressure that can withstand being pressed strongly by hand. AGCL powder is composed entirely of naturally derived materials, showing a hemolysis rate of less than 3％, a cell viability rate of over 99％, and an antibacterial effect of 99.9％, making it safe even when in contact with blood. In animal experiments, excellent tissue regeneration effects such as rapid wound recovery and promotion of blood vessel and collagen regeneration were confirmed. In surgical liver injury experiments, the amount of bleeding and hemostasis time were significantly reduced compared to commercial hemostatic agents, and liver function recovered to normal levels two weeks after surgery. No abnormal findings were observed in systemic toxicity evaluations. In particular, this hemostatic agent maintains its performance for two years even in room temperature and high humidity environments, possessing the advantage of being ready for immediate use in harsh environments such as military operation sites or disaster areas. Although this research is an advanced new material technology developed with national defense purposes in mind, it has great potential for application throughout emergency medicine, including disaster sites, developing countries, and medically underserved areas. It is evaluated as a representative spin-off case＊ where national defense science and technology expanded to the private sector, as it is capable of everything from emergency treatment on the battlefield to internal surgical hemostasis. ＊Spin-off case: Expanding or transferring national defense science and technology for use in the private sector. Examples include computers, GPS, microwave ovens, etc. < Validation of efficacy in wounds through animal experiments / Validation of efficacy in a liver surgery model > This study was recognized for its scientific innovation and national defense utility simultaneously, winning the 2025 KAIST Q-Day President's Award and the Minister of National Defense Award at the 2024 KAIST-KNDU National Defense Academic Conference. Ph.D candidate Kyusoon Park (Army Major), who participated in the research, stated, “The core of modern warfare is minimizing the loss of human life,” and added, “I started the research with a sense of mission to save even one more soldier.” He continued, “I hope this technology will be used as a life-saving technology in both national defense and private medical fields.” This research, in which KAIST PhD student Kyusoon Park and Ph.D candidate Youngju Son participated as lead authors and was guided by Professor Steve Park and Professor Sangyong Jon, was published online on October 28, 2025, in the international academic journal in the field of chemistry/materials engineering, Advanced Functional Materials (IF 19.0). ※ Paper Title: An Ionic Gelation Powder for Ultrafast Hemostasis and Accelerated Wound Healing, DOI: 10.1002/adfm.202523910 Meanwhile, this research was conducted with the support of the National Research Foundation of Korea (NRF)."      

Turning PC and Mobile Devices into AI Infrastructu..

  < (From left) KAIST School of Electrical Engineering: Dr. Jinwoo Park, M.S candidate Seunggeun Cho, and Professor Dongsu Han >       Until now, AI services based on Large Language Models (LLMs) have mostly relied on expensive data center GPUs. This has resulted in high operational costs and created a significant barrier to entry for utilizing AI technology. A research team at KAIST has developed a technology that reduces reliance on expensive data center GPUs by utilizing affordable, everyday GPUs to provide AI services at a much lower cost. On December 28th, KAIST announced that a research team led by Professor Dongsu Han from the School of Electrical Engineering developed 'SpecEdge,' a new technology that significantly lowers LLM infrastructure costs by utilizing affordable, consumer-grade GPUs widely available outside of data centers. SpecEdge is a system where data center GPUs and "edge GPUs"—found in personal PCs or small servers—collaborate to form an LLM inference infrastructure. By applying this technology, the team successfully reduced the cost per token (the smallest unit of text generated by AI) by approximately 67.6％ compared to methods using only data center GPUs. To achieve this, the research team utilized a method called 'Speculative Decoding.' In this process, a small language model placed on the edge GPU quickly generates a high-probability token sequence (a series of words or word fragments). Then, the large-scale language model in the data center verifies this sequence in batches. During this process, the edge GPU continues to generate words without waiting for the server's response, simultaneously increasing LLM inference speed and infrastructure efficiency. < Figure 1. Language data flow diagram of the developed SpecEdge >  < Figure 2. Detailed computation time reduction method of SpecEdge >  < Figure 3. Illustration of efficient batching of verification requests from multiple edge GPUs on the server GPU within SpecEdge > Compared to performing speculative decoding solely on data center GPUs, SpecEdge improved cost efficiency by 1.91 times and server throughput by 2.22 times. Notably, the technology was confirmed to work seamlessly even under standard internet speeds, meaning it can be immediately applied to real-world services without requiring a specialized network environment. Furthermore, the server is designed to efficiently process verification requests from multiple edge GPUs, allowing it to handle more simultaneous requests without GPU idle time. This has realized an LLM serving infrastructure structure that utilizes data center resources more effectively. This research presents a new possibility for distributing LLM computations—which were previously concentrated in data centers—to the edge, thereby reducing infrastructure costs and increasing accessibility. In the future, as this expands to various edge devices such as smartphones, personal computers, and Neural Processing Units (NPUs), high-quality AI services are expected to become available to a broader range of users. < Figure 4. Conceptual comparison of the developed SpecEdge vs. conventional methods > Professor Dongsu Han, who led the research, stated, "Our goal is to utilize edge resources around the user, beyond the data center, as part of the LLM infrastructure. Through this, we aim to lower AI service costs and create an environment where anyone can utilize high-quality AI." Dr. Jinwoo Park and M.S candidate Seunggeun Cho from KAIST participated in this study. The research results were presented as a 'Spotlight' (top 3.2％ of papers, with a 24.52％ acceptance rate) at the NeurIPS (Neural Information Processing Systems) conference, the world's most prestigious academic conference in the field of AI, held in San Diego from December 2nd to 7th. Paper Title: SpecEdge: Scalable Edge-Assisted Serving Framework for Interactive LLMs Paper Links: NeurIPS Link, arXiv Link This research was supported by the Institute of Information & Communications Technology Planning & Evaluation (IITP) under the project 'Development of 6G System Technology to Support AI-Native Application Services.'      

KAIST Researchers First in the World to Identify S..

  <Photo 1. (From left) Ph.D. candidates Mingyoo Song and Jaehan Kim, Professor Sooel Son, (Top right) Professor Seungwon Shin, Lead Researcher Seung Ho Na>   Most major commercial Large Language Models (LLMs), such as Google’s Gemini, utilize a Mixture-of-Experts (MoE) structure. This architecture enhances efficiency by dynamically selecting and using multiple "small AI models (Expert AIs)" depending on input queries . However, KAIST research team has revealed for the first time in the world that this very structure can actually become a new security threat. A joint research team led by Professor Seungwon Shin (School of Electrical Engineering) and Professor Sooel Son (School of Computing) announced on December 26th that they have identified an attack technique that can seriously compromise the safety of LLMs by exploiting the MoE structure. For this research, they received the Distinguished Paper Award at ACSAC 2025, one of the most prestigious international conferences in the field of information security. ACSAC (Annual Computer Security Applications Conference) is among the most influential international academic conferences in security. This year, only two papers out of all submissions were selected as Distinguished Papers. It is highly unusual for a domestic Korean research team to achieve such a feat in the field of AI security. In this study, the team systematically analyzed the fundamental security vulnerabilities of the MoE structure. In particular, they demonstrated that even if an attacker does not have direct access to the internal structure of a commercial LLM, the entire model can be induced to generate dangerous responses if just one maliciously manipulated "Expert Model" is distributed through open-source channels and integrated into the system. <Figure 1. Conceptual diagram of the attack technology proposed by the research team.> To put it simply: even if there is only one "malicious expert" mixed among normal AI experts, that specific expert may be repeatedly selected for processing harmful queries, causing the overall safety of the AI to collapse. A particularly dangerous factor highlighted was that this process causes almost no degradation in model performance, making the problem extremely difficult to detect in advance. Experimental results showed that the attack technique proposed by the research team could increase the harmful response rate from 0％ to up to 80％. They confirmed that the safety of the entire model significantly deteriorates even if only one out of many experts is "infected." This research is highly significant as it presents the first new security threat that can occur in the rapidly expanding global open-source-based LLM development environment. Simultaneously, it suggests that verifying the "source and safety of individual  expert models" is now essential—not just performance—during the AI model development process. Professors Seungwon Shin and Sooel Son stated, "Through this study, we have empirically confirmed that the MoE structure, which is spreading rapidly for the sake of efficiency, can become a new security threat. This award is a meaningful achievement that recognizes the importance of AI security on an international level." The study involved Ph.D. candidates Jaehan Kim and Mingyoo Song, Dr. Seung Ho Na (currently at Samsung Electronics), Professor Seungwon Shin, and Professor Sooel Son. The results were presented at ACSAC in Hawaii, USA, on December 12, 2025. <Figure 2. Photo of the Distinguished Paper Award certificate> Paper Title: MoEvil: Poisoning Experts to Compromise the Safety of Mixture-of-Experts LLMs Paper File: https://jaehanwork.github.io/files/moevil.pdf GitHub (Open Source): https://github.com/jaehanwork/MoEvil This research was supported by the Korea Internet & Security Agency (KISA) and the Institute of Information & Communications Technology Planning & Evaluation (IITP) under the Ministry of Science and ICT.        

Breakthrough in Intractable Intestinal Disease Tre..

  < (From left) Professor Sung Gap Im (KAIST), Dr. Seonghyeon Park (KAIST), M.S candidate Sang Yu Sun (KAIST), Dr. Mi-Young Son (KRIBB), (Top right) Dr. Tae Geol Lee (KRISS), Dr. Jin Gyeong Son (KRISS) >   Intestinal Stem Cells (ISCs) derived from a patient's own cells have garnered significant attention as a new alternative for treating intractable intestinal diseases due to their low risk of rejection. However, clinical application has been limited by safety and regulatory issues arising from conventional culture methods that rely on animal-derived components (xenogeneic components). A KAIST research team has developed an advanced culture technology that stably grows ISCs without animal components while simultaneously enhancing their migration to damaged tissues and regenerative capabilities. KAIST announced on December 23rd that a joint research team—led by Professor Sung Gap Im from the Department of Chemical and Biomolecular Engineering, Dr. Tae Geol Lee from the Nano-Bio Measurement Group at the Korea Research Institute of Standards and Science and Dr. Mi-Young Son from the Stem Cell Convergence Research Center at the Korea Research Institute of Bioscience and Biotechnology  has developed a polymer-based culture platform that dramatically improves the migration and regeneration of ISCs in a xenogeneic-free environment. To overcome obstacles in the clinical application of stem cell therapies—such as the risk of virus transmission to patients when using substances derived from mouse fibroblasts or Matrigel—the joint research team developed "PLUS" (Polymer-coated Ultra-stable Surface). This polymer-based culture surface technology functions effectively without any animal-derived materials. < Figure 1. Precise control of polymer coating and surface modification via initiated Chemical Vapor Deposition (iCVD) process > PLUS is a synthetic polymer surface coated via a vapor deposition method. By precisely controlling surface energy and chemical composition, it significantly enhances the adhesion and mass-culture efficiency of ISCs. Notably, it maintains identical culture performance even after being stored at room temperature for three years, securing industrial scalability and storage convenience for stem cell therapeutics. Through proteomics analysis＊, the research team identified that the expression of proteins related to cytoskeletal reorganization significantly increased in ISCs cultured on the PLUS environment. Proteomics Analysis: A method used to simultaneously analyze the types and quantitative changes of all proteins present within a cell or tissue. Specifically, the team confirmed that increased expression of cytoskeleton-binding and actin-binding proteins leads to a stable restructuring of the internal cellular architecture. This provides the power source for stem cells to move faster and more actively across the substrate. < Figure 2. Elucidation of the mechanism for enhanced ISC migration through precision proteomics analysis > Real-time observations using holotomography microscopy revealed that ISCs cultured on PLUS exhibited a migration speed approximately twice as fast as those on conventional surfaces. Furthermore, in a damaged tissue model, the cells demonstrated outstanding regenerative performance, repairing more than half of the damage within a single week. This proves that PLUS activates the cytoskeletal activity of stem cells, thereby boosting their practical tissue regeneration capabilities. The newly developed PLUS culture platform is evaluated as a technology that will significantly enhance the safety, mass production, and clinical feasibility of ISCs derived from human pluripotent stem cells (hPSCs). By elucidating the mechanism that simultaneously strengthens the survival, migration, and regeneration of stem cells in a xenogeneic-free environment, the team has established a foundation to fundamentally resolve safety, regulatory, and productivity issues in stem cell therapy. Professor Sung Gap Im of KAIST stated, "This research provides a synthetic culture platform that eliminates the dependence on xenogeneic components—which has hindered the clinical application of stem cell therapies—while maximizing the migration and regenerative capacity of stem cells. It will serve as a catalyst for a paradigm shift in the field of regenerative medicine." Dr. Seonghyeon Park (KAIST), Sang Yu Sun (KAIST), and Dr. Jin Gyeong Son (KRISS) participated as first authors. The research findings were published online on November 26th in Advanced Materials, the leading academic journal in materials science. Paper Title: Tailored Xenogeneic-Free Polymer Surface Promotes Dynamic Migration of Intestinal Stem Cells DOI: 10.1002/adma.202513371 This research was conducted with support from the Ministry of Science and ICT, the Ministry of SMEs and Startups, the National Research Foundation of Korea, the National Council of Science and Technology Research, KRISS, KRIBB, and the National NanoFab Center.    

KAIST, AI judges manufacturing beyond craftsmanshi..

  <(From Left) M.S candidate Inhyo Lee, Ph.D candidate Heekyu Kim, Ph.D candidate joonyoung Kim, Professor Seunghwa Ryu>   Most of the plastic products we use are made through injection molding, a process in which molten plastic is injected into a mold to mass-produce identical items. However, even slight changes in conditions can lead to defects, so the process has long relied on the intuition of highly skilled workers. Now, KAIST researchers have proposed an AI-based solution that autonomously optimizes processes and transfers manufacturing knowledge, addressing concerns that expertise could be lost due to the retirement of skilled workers and the increase in foreign labor. KAIST (President Kwang Hyung Lee) announced on the 22nd of December that a research team led by Professor Seunghwa Ryu from the Department of Mechanical Engineering · InnoCORE PRISM-AI Center has, for the first time in the world, developed generative AI technology that autonomously optimizes injection molding processes, along with an LLM-based knowledge transfer system that makes on-site expertise accessible to anyone. The team also reported that these achievements were published consecutively in an internationally renowned journal. The first achievement is a generative AI–based process inference technology that automatically infers optimal process conditions based on environmental changes or quality requirements. Previously, whenever temperature, humidity, or desired quality levels changed, skilled workers had to rely on trial and error to readjust conditions. The research team implemented a diffusion model–based approach that reverse-engineers process conditions satisfying target quality requirements, using environmental data and process parameters collected over several months from an actual injection molding factory. In addition, the team built a surrogate model that substitutes for actual production, enabling quality prediction without running the real process. As a result, they achieved an error rate of just 1.63％, significantly lower than the 23~44％ error rates of representative existing technologies such as GAN＊ and VAE＊＊ models traditionally used for process prediction. Experiments applying the AI-generated conditions to real processes confirmed successful production of acceptable products, demonstrating practical applicability.  ＊GAN (Generative Adversarial Network): a method in which two AI models compete with each other to generate data  ＊＊VAE (Variational Autoencoder): a method that compresses and learns common patterns in data and then reconstructs them <Figure 1. Generative AI–Based Process Reasoning Technology> The second achievement is the IM-Chat, an LLM-based knowledge transfer system designed to address skilled worker retirement and multilingual work environments. IM-Chat is a multi-agent AI system that combines large language models (LLMs) with retrieval-augmented generation (RAG), serving as an AI assistant for manufacturing sites by providing appropriate solutions to problems encountered by novice or foreign workers. When a worker asks a question in natural language, the AI understands it and, if necessary, automatically calls the generative process inference AI, simultaneously providing optimal process condition calculations along with relevant standards and background explanations. For example, when asked, “What is the appropriate injection pressure when the factory humidity is 43.5％?”, the AI calculates the optimal condition and presents the supporting manual references as well. With support for multilingual interfaces, foreign workers can receive the same level of decision-making support. This research is regarded as a core manufacturing AI transformation (AX) technology that can be extended beyond injection molding to molds, presses, extrusion, 3D printing, batteries, bio-manufacturing, and other industries. In particular, the work is significant in that it presents a paradigm for autonomous manufacturing AI, integrating generative AI and LLM agents through a Tool-Calling approach＊, enabling AI to make its own judgments and invoke necessary functions.  ＊Tool-Calling approach: a method in which AI autonomously calls and uses the functions or programs required for a given situation <Figure 2. Large Language Model–Based Multilingual Knowledge Transfer Multi-Agent IM-Chat> <Figure 3. Example of Operation of the Large Language Model (LLM)–Based Multilingual Knowledge Transfer Multi-Agent IM-Chat> <Figure 4. Illustration of the Application of an LLM-Based Multilingual Knowledge Transfer Multi-Agent IM-Chat (AI-Generated)> Professor Seunghwa Ryu explained, “This is a case where we addressed fundamental problems in manufacturing in a data-driven way by combining AI that autonomously optimizes processes with LLMs that make on-site knowledge accessible to anyone,” adding, “We will continue expanding this approach to various manufacturing processes to accelerate intelligence and autonomy across the industry.” This research involved doctoral candidates Junhyeong Lee, Joon-Young Kim, and Heekyu Kim from the Department of Mechanical Engineering as co–first authors, with Professor Seunghwa Ryu as the corresponding author. The results were published consecutively in the April and December issues of Journal of Manufacturing Systems (JCR 1/69, IF 14.2), the world’s top-ranked international journal in engineering and industrial fields.  ※ Paper 1: “Development of an Injection Molding Production Condition Inference System Based on Diffusion Model,” DOI: https://doi.org/10.1016/j.jmsy.2025.01.008  ※ Paper 2: “IM-Chat: A multi-agent LLM framework integrating tool-calling and diffusion modeling for knowledge transfer in injection molding industry,” DOI: https://doi.org/10.1016/j.jmsy.2025.11.007 This research was supported by the Ministry of Science and ICT, the Ministry of SMEs and Startups, and the Ministry of Trade, Industry and Energy.  

Where did this fish come from？ Securing World-Clas..

  < (From left) KAIST Ph.D. candidate Hyeontaek Hwang, Research Professor Yalew Kidane, Senior Researcher Young-jong Lee, Researcher Geon-woo Park, and (Top) Professor Daeyoung Kim >   When buying seafood at a supermarket, you may have wondered where the fish was caught and what process it went through to reach your dinner table. However, due to complex distribution processes, it has been difficult to transparently track that path. KAIST’s research team has developed a digital technology that solves this problem, allowing the movement path of seafood to be checked at a glance based on international standards recognized worldwide. KAIST announced on December 19th that "OLIOPASS," a GS1 international standard-based digital transformation solution developed by Director Daeyoung Kim (Professor, School of Computing) of the KAIST Auto-ID Labs Busan Innovation Center, has passed the rigorous performance verification of the GDST (Global Dialogue on Seafood Traceability). It is the first in Korea to obtain the "GDST Capable Solution" certification. < (Left) GDST Global Certification Logo, (Right) KAIST OLIOPASS Platform Logo > Only 13 technologies worldwide have received this GDST certification. Among them, only 7 entities, including KAIST, support "Full Chain" traceability technology, which manages the entire process from production and processing to distribution and sales. The GDST is an international organization established in 2015 at the suggestion of the World Economic Forum (WEF). It helps record and share information on all seafood movement processes digitally, according to the GS1 international standard agreed upon by the global community. This can be compared to creating a "common language for the supply chain" used worldwide. The GDST is a global standard system that increases the reliability of seafood history information by defining Key Data Elements (KDEs) that must be recorded during the movement of seafood and Critical Tracking Events (CTEs) that define when, where, and what moved, based on international standards. As major food distribution companies in the United States and Europe have recently begun requiring GDST compliance, this standard is becoming a de facto essential requirement for entering the global market. Since 2019, KAIST has participated as a founding member of GDST and has played a key role in designing seafood traceability models and system-to-system information interoperability. In particular, with the U.S. Food and Drug Administration (FDA) announcing the mandatory enforcement of food traceability (FSMA 204) starting in July 2028, this certification is significant as it secures a technical solution for domestic companies to meet global market regulations. OLIOPASS, which received certification on November 5th, is a digital traceability platform that combines KAIST's IoT technology with international standards (GS1 EPCIS 2.0, GS1 Digital Link). It records and shares movement information of various products and assets in a standardized language and utilizes blockchain technology to fundamentally prevent forgery or alteration. Even if systems differ between companies, history data is seamlessly linked. Furthermore, OLIOPASS is designed as an "AI-ready data" infrastructure, allowing for the easy application of next-generation AI technologies such as Large Multimodal Models (LMM), AI agents, knowledge graphs, and ontologies. This allows it to serve as a platform that supports both digital and AI transformation beyond simple history management. Daeyoung Kim, Director of the KAIST Auto-ID Labs Busan Innovation Center, stated, "This certification is an international recognition of our capability in reliable data technology across the global supply chain. We will expand OLIOPASS beyond seafood and food into various fields such as pharmaceuticals, logistics, defense, and smart cities, ensuring KAIST’s technology grows into a platform used by the world." ※ Related Link: https://thegdst.org/verified-gdst-capable-solutions/ < List of Certified Organizations >    

AI Gets a Private Tutor, Learning Human Preference..

  < Professor Junmo Kim and Ph.D. candidate Minchan Kwon, School of Electrical Engineering >   No matter how much data they learn, why do Artificial Intelligence (AI) models often miss the mark on human intent? Conventional "comparison learning," designed to help AI understand human preferences, has frequently led to confusion rather than clarity. A KAIST research team has now presented a new learning solution that allows AI to accurately learn human preferences even with limited data by assigning it a "private tutor." On December17th, a research team led by Professor Junmo Kim of KAIST School of Electrical Engineering announced the development of "TVKD" (Teacher Value-based Knowledge Distillation), a reinforcement learning framework that significantly improves data efficiency and learning stability while effectively reflecting human preferences. Existing AI training methods typically rely on collecting massive amounts of "preference comparison" data—simple structures like "A is better than B." However, this approach requires vast datasets and often causes the AI to become confused in ambiguous situations where the distinction is unclear. To solve this problem, the research team proposed a method in which a ‘Teacher model’ that has first deeply understood human preferences delivers only the core information to a ‘Student model.’ This can be compared to a private tutor who organizes and teaches complex content, and the research team named this ‘Preference Distillation.’ The biggest feature of this technology is that instead of simply imitating ‘good or bad,’ it is designed so that the teacher model learns a ‘Value Function’ that numerically judges how valuable each situation is, and then delivers this to the student model. Through this, the AI can learn by making comprehensive judgments about ‘why this choice is better’ rather than fragmentary comparisons, even in ambiguous situations.   < Conceptual diagram of TVKD: After teaching the human preference dataset to the teacher model, learning proceeds by delivering the teacher's information and the dataset to the student model > The core of this technology is twofold. First, by reflecting value judgments that consider the entire context into the student model, learning that understands the overall flow rather than fragmentary answers has become possible. Second, a technique was introduced to adjust learning importance according to the reliability of preference data. Clear data is significantly reflected in learning, while the influence of ambiguous or noisy data is reduced, allowing the AI to learn stably even in realistic environments. As a result of the research team applying this technology to various AI models and conducting experiments, it showed more accurate and stable performance than methods previously known to have the best performance. In particular, it recorded achievements that stably outperformed existing top technologies in major evaluation indices such as MT-Bench and AlpacaEval. Professor Junmo Kim said, “In reality, human preference data is not always sufficient or perfect,” and added, “This technology will allow AI to learn consistently even under such constraints, so it will be highly practical in various fields.” < Performance comparison results for each task of MT-Bench. It can be confirmed that the proposed TVKD framework records generally higher scores than existing methods. > < Visualization results of the Shaping term. The top tokens (converted into words) judged as important by the teacher model within the response are displayed in red, intuitively showing which tokens have a greater influence during the value-based alignment process. > Ph.D. candidate Minchan Kwon from our university’s School of Electrical Engineering participated as the first author, and the research results were accepted at ‘NeurIPS 2025’, the most prestigious international conference in the field of artificial intelligence. The research was presented at a poster session on December 3, 2025 (US Pacific Time). ※ Paper Title: Preference Distillation via Value based Reinforcement Learning, DOI: https://doi.org/10.48550/arXiv.2509.16965 Meanwhile, this research was carried out with support from the Information & Communications Technology Planning & Evaluation (IITP) funded by the government (Ministry of Science and ICT) in 2024 (No. RS-2024-00439020, Development of Sustainable Real-time Multimodal Interactive Generative AI, SW Star Lab).  

Harry Potter–Style ‘Moving Invisibility Cloak’ Tec..

  <(Top row, left) Ph.D candidate Hyeonseung Lee, Professor Wonho Choe, (Second row, left) Professor Hyoungsoo Kim, Professor Sanghoo Park,(Top) First author Dr. Jeongsu Pyeon>     What do Harry Potter’s invisibility cloak and stealth fighter jets that evade radar have in common? They both make objects invisible despite their physical presence. Building upon this concept, our research team has taken it one step further by developing a “smart invisibility cloak” like technology that hides electromagnetic waves even better as it stretches and moves. This technology is expected to open new possibilities for moving robots, body-mounted wearable devices, and next-generation stealth technologies. On December 16th, research teams led by Professor Hyoungsoo Kim of the Department of Mechanical Engineering and Professor Sanghoo Park of the Department of Nuclear and Quantum Engineering from KAIST announced that they have developed a core enabling technology for next-generation stretchable cloaking＊ based on Liquid Metal Composite Ink (LMCP), which can absorb, modulate, and shield electromagnetic waves. ＊ Cloaking: A technology that makes an object appear as if it does not exist to detection equipment such as radar or sensors, even though it is physically present. To realize cloaking technology, it is necessary to freely control light or electromagnetic waves on the surface of an object. However, conventional metallic materials are rigid and do not stretch well, and when forcibly stretched, they easily break. For this reason, there have been significant difficulties in applying such materials to body-conforming electronic devices or robots that freely change shape. The liquid metal composite ink developed by the research team maintains electrical conductivity even when stretched up to 12 times its original length (1200％), and it demonstrated high stability with little oxidation or performance degradation even after being left in air for nearly a year. Unlike conventional metals, this ink is rubber-like and soft while fully retaining metallic functionality. These properties are possible because, during the drying process, liquid metal particles inside the ink spontaneously connect with one another to form a mesh-like metallic network structure. This structure functions as a “metamaterial”—an artificial structure in which extremely small patterns are repeatedly printed using ink so that electromagnetic waves interact with the structure in a designed manner. As a result, the material simultaneously exhibits liquid-like flexibility and metal-like robustness. The fabrication process is also simple. Without complex procedures such as high-temperature sintering or laser processing, the ink can be printed using a printer or applied with a brush and then simply dried. In addition, common drying issues such as stains or cracking do not occur, enabling smooth and uniform metal patterns. To verify the performance of the ink, the research team became the first in the world to fabricate a “stretchable metamaterial absorber” whose electromagnetic wave absorption characteristics change depending on the degree of stretching. Simply stretching the rubber-like substrate after printing patterns with the ink changes the type (frequency band) of electromagnetic waves that are absorbed. This demonstrates the potential for cloaking technology that can more effectively hide objects from radar or communication signals depending on the situation. <Figure. Comparison of LMCP ink properties, printing process applicability, mechanical/electrical performance, and versatility on various substrates. (a) Comparison results regarding surface tension, viscosity, wettability, and post-processing requirements between conventional liquid metal-based inks and the LMCP ink in this study. The results demonstrate that LMCP ink possesses the advantage of requiring no post-processing while maintaining relatively high viscosity and excellent wettability. (Right radar chart: Qualitative comparison of key performance indicators, including electrical conductivity, surface tension, viscosity, wettability, and post-processing requirements). (b) Various printing methods based on the self-sintering characteristics of LMCP ink: nozzle-based direct writing, brushing, patterning using shadow masks and doctor blade processes, and large-area electrode fabrication via the roll-to-roll method. (c) Stretchability and electrical stability of LMCP electrodes. Results show resistance changes when samples are stretched from 0％ to 1200％, and stable operation is confirmed under 0％–500％ strain through a 3 V LED driving experiment. (d) Examples of various patterns and devices fabricated using LMCP ink. Applicable structures are presented, including large-area uniform coating, precise grid patterns, crack-free metal paths, LED circuits operating under tension, and stretchable spiral electrodes>   (e) Examples demonstrating stable printing of LMCP ink on various substrates (SIR, NBR, PVC, PET, WPU, PDMS, Latex), indicating excellent pattern reproducibility and adhesion regardless of the substrate type>   This technology is evaluated as a groundbreaking electronic material technology that simultaneously satisfies stretchability, electrical conductivity, long-term stability, process simplicity, and electromagnetic wave control functionality. Professor Hyoungsoo Kim stated, “We have made it possible to implement electromagnetic wave functionality using only printing processes without complex equipment,” adding, “This technology is expected to be utilized in various future technologies such as robotic skin, body-mounted wearable devices, and radar stealth technologies in the defense sector.” This research was recognized as an important fundamental technology in the field of next-generation electronic materials and was published in the October 2025 issue of the international Wiley journal Small on October 16, where it was selected as a cover article. Paper title: J. Pyeon, H. Lee, W. Choe, S. Park, H. Kim, “Versatile Liquid Metal Composite Inks for Printable, Durable, and Ultra-Stretchable Electronics,” Small 2501829 (2025) DOI: https://doi.org/10.1002/smll.202501829 Author information: First author: Dr. Jeongsu Pyeon Co-authors: Doctoral candidate Hyeonseung Lee, Professor Wonho Choe Corresponding authors: Professor Hyoungsoo Kim, Professor Sanghoo Park This work was supported by the National Research Foundation of Korea’s Mid-Career Research Program (MSIT: 2021R1A2C2007835) and the KAIST UP Program.   < Selected as the cover article of the October 2025 issue of the international journal Small > < Invisibility cloak technology image (AI-generated image) >

Octopus-Inspired 3D Micro-LEDs Pave the Way for Se..

  <(From Left) Professor Keon Jae Lee, Professor Tae-Hyuk Kwon, Ph.D candidate Min Seo Kim, Dr. Jae Hee Lee, Dr. Chae Gyu Lee>   -KAIST and UNIST Researchers Develop Shape-Morphing Device to Overcome Pancreatic Tumor Microenvironment Barriers   Conventional pancreatic cancer treatments face a critical hurdle due to the dense tumor microenvironment (TME). This biological barrier surrounds the tumor, severely limiting the infiltration of chemotherapy agents and immune cells. While photodynamic therapy (PDT) offers a promising alternative, existing external light sources, such as lasers, fail to penetrate deep tissues effectively and pose risks of thermal damage and inflammation to healthy organs   To address these challenges, Professor Keon Jae Lee’s team at KAIST, in collaboration with Professor Tae-Hyuk Kwon at UNIST, developed an implantable, shape-morphing 3D micro-LED device capable of effectively delivering light to deep tissues. The key technology lies in the device’s flexible, octopus-like architecture, which allows it to wrap around the entire pancreatic tumor. This mechanical compliance ensures uniform light delivery to the tumor despite the tumor’s physiological expansion or contraction, enabling continuous, low intensity photostimulation that precisely targets cancer cells while preserving normal tissue.   In in-vivo experiments involving mouse models, the device demonstrated remarkable therapeutic efficacy. Within just three days, tumor fibrous tissue was reduced by 64％, and the pancreatic tissue successfully reverted to normal tissue, overcoming the limitations of conventional PDT.   Prof. Keon Jae Lee said, "This research presents a new therapeutic paradigm by directly disrupting the tumor microenvironment, the primary obstacle in pancreatic cancer treatment." He added, "We aim to expand this technology into a smart platform integrated with artificial intelligence (AI) for real-time tumor monitoring and personalized treatment. We are currently seeking partners to advance clinical trials and commercialization for human application."   <Overall concept of 3D Shape-morphing micro-LEDs (SMLEDs). The 3D long-term, low-intensity photodynamic therapy (PDT) system attaches to the pancreatic surface, ensuring stable and continuous light delivery. Initially maintaining a 2D structure, the system morphs into a 3D structure upon implantation to conform to the shape of the pancreas. In in vivo experiments, the device maintained stable adhesion without detachment for four weeks and reduced the pancreatic tumor size by 64％.> Professor Tae-Hyuk Kwon commented, "While phototherapy is effective for selective cancer treatment, conventional technologies have been limited by the challenges of delivering light to deep tissues and developing suitable photosensitizers." He added, "Building on this breakthrough, we aim to expand effective immune-based therapeutic strategies for targeting intractable cancers."   <Cover Image. The 3D long-term, low-intensity photodynamic therapy (PDT) system, developed by Professor Keon Jae Lee's team at the Department of Materials Science and Engineering at KAIST, was featured as the cover article of the international journal Advanced Materials> The result, titled "Deeply Implantable, Shape-Morphing, 3D MicroLEDs for Pancreatic Cancer Therapy," was featured as the cover article in Advanced Materials (Volume 37) on December 10, 2025.