Kaist

KAIST Professor Sooel Son Selected for Microsoft F..

  < Professor Sooel Son, School of Computing >     KAIST (President Choongsik Bae) announced on the 28th of July that Professor Sooel Son has been selected as the only researcher in Korea to receive funding from Microsoft for research on artificial intelligence safety and security. The funding was awarded through the External Red Team Alliance (EXTRA), a new program established by Microsoft’s AI Red Team, which examines the safety and security vulnerabilities of AI systems. EXTRA is a global initiative designed to strengthen AI safety and security research capabilities by supporting researchers at universities and technology experts around the world. Microsoft noted that most AI safety testing is still conducted internally by individual companies or organizations. However, assessing the major risks posed by increasingly advanced AI systems requires expertise across a broad range of fields, including cybersecurity, multilingual environments, regional and cultural contexts, AI alignment, and potential misuse. EXTRA was launched in recognition of the difficulty a single internal organization faces when comprehensively evaluating these diverse risks. Through the program, Microsoft will provide KAIST with an unrestricted gift of USD 25,000, approximately KRW 37 million, with no prescribed project period, to support research related to AI safety, security, alignment, and responsible AI development. The funding will be used to support Professor Son’s research team in its work on AI security and safety. More than a dozen universities across six continents are participating in EXTRA, with KAIST being the only university selected from Korea. Through the program, Microsoft plans to expand the ecosystem for independent AI safety research and strengthen collaboration between academia and industry. Professor Son’s research team has been conducting research on the security and privacy of AI systems that use machine-learning models and large language models. In particular, the team analyzes adversarial attacks against deep neural networks and language models—including model extraction, membership inference, personal information extraction, model inversion, machine unlearning, and prompt injection—and develops defense methodologies to assess and improve model safety. Building on these technologies, the team is also focusing on establishing systematic defense methodologies that enable the safe and trustworthy deployment of agentic AI systems operating in real-world service environments, including web agents and agentic browsers. “As AI systems rapidly spread throughout society, research that verifies the security and reliability of increasingly advanced AI is becoming more important,” said Professor Son. “Our participation in Microsoft AI Red Team’s EXTRA program will provide an opportunity to further advance our research on the safe development and use of AI systems.” “AI safety research has never been more important, and universities have a critical role to play in advancing the field,” said Ram Shankar Siva Kumar, who leads the Microsoft AI Red Team. “Through EXTRA, we aim to support researchers working to deepen our understanding of how increasingly powerful AI systems can be evaluated, protected, and governed responsibly.” “Competition in AI technology is expanding beyond performance to encompass safety and trustworthiness,” said KAIST President Choongsik Bae. “KAIST’s participation as the only Korean research institution in Microsoft’s global AI safety research network is a meaningful achievement that demonstrates Korea’s competitiveness in AI research. We will continue to lead the advancement of responsible AI technologies that everyone can trust and use by pursuing world-class research in AI safety and security.”  

KAIST Makes Cancer Cells Send Out Their Own Danger..

  < Prof. Yeu-Chun Kim (KAIST), Prof. Yong-Kyu Lee (Korea National University of Transportation), Dr. A-Reum Yoon (Hanyang University), and Dr. Soo-Sam Lee (KAIST) >     Cancer cells survive by hiding from the immune system's surveillance. A KAIST research team has developed a new anticancer platform that makes cancer cells send out their own danger signal—prompting immune cells to attack—while simultaneously delivering gene therapy. The approach is expected to offer a new treatment strategy that combines cancer immunotherapy and gene therapy in a single nanoparticle. Immunogenic cell death (ICD) is a process in which dying cancer cells send danger signals to nearby immune cells, prompting them to attack. A polypeptide is a polymer made of a long chain of amino acids. KAIST (President Choongsik Bae) announced on 28th of July that a team led by Professor Yeu-Chun Kim from the Department of Chemical and Biomolecular Engineering has developed a "helical polypeptide nanoparticle" platform that induces severe stress inside cancer cells to trigger immunogenic cell death, while also delivering a range of gene therapeutics into the cells. The body's immune cells effectively eliminate external invaders such as viruses and bacteria, but cancer cells evade immune surveillance through a variety of immune-escape strategies. This failure of immune cells to recognize cancer cells as a threat has long been one of the biggest limitations in cancer treatment. Recently, researchers have been actively exploring the use of nanoparticles to deliver drugs and genes to cancer cells and activate immune responses. However, it has not been clearly established which properties of nanomaterials actually induce cellular stress and activate antitumor immune responses. By comparing and analyzing a range of nanoparticles, the team confirmed that not just the chemical composition, but the helical, coiled shape of the nanomaterial is a key factor determining therapeutic efficacy. In particular, when a positively charged quaternary amine—a chemical structure that binds readily to cell membranes—was combined with a helical structure, the particle could penetrate the cell membrane like a screw and enter cancer cells with ease. By contrast, particles with the same chemical composition but lacking the helical coil barely entered cells at all and failed to induce an immune response. The helical nanoparticles developed by the team preferentially seek out and penetrate cancer cells, which have different membrane electrical properties from normal cells. Once inside, the particles disrupt the membranes of mitochondria—the cell's energy-producing organelles—and other organelles, subjecting the cancer cell to severe stress. Under this extreme stress, the dying cancer cell releases damage-associated molecular patterns (DAMPs)—distress signals indicating "a dangerous cell is here"—into the surrounding environment. Immune cells that detect these signals recognize the previously hidden cancer cell as a threat and begin their attack. In effect, the cancer cell is made to broadcast its own location to the immune system. The nanoparticle does more than trigger an immune response—it also functions as a carrier for gene therapeutics. Messenger RNA (mRNA), which carries the genetic information for protein synthesis, and small interfering RNA (siRNA), which suppresses the expression of specific genes, are both typically difficult to deliver into cells. The team's nanoparticles, however, delivered these molecules effectively into the cytoplasm. The researchers also introduced guanidinium, a chemical functional group that binds strongly to genetic material, at an optimized ratio, enabling the particles to remain stable in the bloodstream while delivering gene therapeutics effectively. In mouse models of melanoma and colorectal cancer, the team loaded the helical nanoparticles with siRNA targeting PD-L1 (Programmed Death-Ligand 1), an immune-evasion protein, and administered them. Tumor growth was suppressed by approximately 70–80％, and a marked increase was observed in cytotoxic T cells—which directly attack cancer cells—infiltrating the tumor, indicating a substantial boost in antitumor immune response. "This study presents a new anticancer platform in which the nanomaterial does more than simply deliver a therapeutic agent—it drives cancer cells to trigger their own immune response," said Professor Yeu-Chun Kim. He added that the platform is expected to contribute to the development of next-generation treatments combining cancer immunotherapy and gene therapy. Dr. Susam Lee, the paper's first author, added, "We showed that it is not just the composition of the nanomaterial but the helical structure itself that is the key factor determining therapeutic efficacy." He said he hopes the findings will serve as a new benchmark for designing next-generation immuno-oncology nanomaterials. The study was published online in Biomaterials, a leading international journal in the field of biomaterials, on May 28, 2026.   Paper title: Helical quaternary amine polypeptide programs membrane stress to drive immunogenic cell death and cytosolic gene delivery for cancer immunotherapy DOI: 10.1016/j.biomaterials.2026.124337   This work was supported by the National Research Foundation of Korea (NRF) grants funded by the Korean government (MSIT), the Biomedical Global Talent Nurturing Program of the Korea Health Industry Development Institute (KHIDI), funded by the Ministry of Health & Welfare, Republic of Korea (RS-2025-25459605 to Susam Lee) and the Korea Basic Science Institute (National research Facilities and Equipment Center).  

KAIST Develops AI That Learns to Theorize the Worl..

  < The research team. From left: Sungjin Ahn, Professor (KAIST School of Electrical Engineering and Computer Science & School of Computing); Doojin Baek, Master >   A KAIST research team has developed a next-generation world model, an internal model an AI builds to understand and predict the world, that learns executable theories from observation alone. KAIST (President Choongsik Bae) announced on the 15th of July that a team led by Professor Sungjin Ahn from the School of Computing has proposed a new learning paradigm called Learning-to-Theorize (L2T), which trains AI to theorize how the world works using only observed information. The team also built the Neural Theorizer (NEO), a neural network-based model that implements this paradigm. The research was selected for an oral presentation at the 43rd International Conference on Machine Learning (ICML 2026), held in Seoul from July 6 to 11, and was presented on July 9. This places it among the top 0.7 percent (168 papers) of the 23,918 total submissions. The paper was also selected for the Best Paper Award at the Compositional Learning Workshop. World models are a foundational technology across robot control, autonomous driving, generative AI, and autonomous agents — AI systems capable of judging and acting on their own. Until now, world models have mainly focused on predicting what happens next. Even when a model predicts the next scene accurately, it doesn't necessarily understand why that change occurred — that is, the underlying principle governing the world. < The award ceremony at the Compositional Learning Workshop, ICML 2026. NEO was selected and honored as the Best Paper Award. >   The team found a solution in how humans learn. Long before children acquire language, they build their own internal theories of how the world works. Applying this view from developmental cognitive science to AI, the researchers built a system that understands the principles behind the world, rather than one that simply predicts the future. The team's proposed L2T framework provides no predetermined answers or rules. Given only a "before" and "after" observation, the AI discovers on its own which rule produced the change. While conventional AI models focus on "guessing what comes next," this approach is built to understand "why the change happened." To implement this, the team also developed Neural Theorizier, NEO. The model discovers reusable primitives hidden within observed transformations and composes them into executable programs. These learned primitives can then be systematically recombined to explain new situations. < Figure 1. Learning to Theorize (L2T) Framework. (a) Training data consists of observation pairs (x, y) generated by unobserved true programs. (b) Under L2T, the model learns to discover reusable primitives (Rotate, Left, Down, and Paint) and to compose them into executable theories. (c) Without L2T, the model instead memorizes entangled composite primitives (e.g., Left-Down) as indecomposed single units. (d) Once the model has learned to theorize, novel phenomena (e.g., Down-Paint-Rotate) can be explained by recombining learned primitives. (e) In contrast, memorized entangled representations fail to generalize to unseen programs. >   For example, NEO independently learns primitives corresponding to basic operations such as rotation, movement to the left or downward, and coloring. Even when presented with a combination it has never encountered during training, such as “move down, then color, then rotate”, it can recombine the primitives it has already learned to explain and solve the new situation. Conventional AI, by contrast, tends to memorize entangled patterns as a single unit, so its performance drops sharply when faced with an unfamiliar combination. Through a range of experiments, the team demonstrated that NEO outperforms existing approaches in compositional generalization, the ability to combine learned basic rules to solve problems never seen before. "It points to a new direction beyond prediction-centric world models, what we call a 'World Theory Model.' We expect this to develop into a core technology across fields including intelligent robots, autonomous agents, and AI that supports scientific discovery." said Professor Sungjin Ahn. Master's students Doojin Baek and Gyubin Lee from the School of Computing served as co-first authors on the study.   Paper title: Learning to Theorize the World from Observation. DOI: https://doi.org/10.48550/arXiv.2605.03413 Authors: Doojin Baek＊, Gyubin Lee＊, Junyeob Baek, Hosung Lee, Sungjin Ahn (＊Co-first authors)   The research was supported by the National Research Foundation of Korea (NRF).      

KAIST Develops Low-Temperature Technique for Growi..

  < The research team. From left: Gayeon Lee, Ph.D. student; Mingyu Jang, student in the integrated M.S.–Ph.D. program; Changhwan Kim, student in the integrated M.S.–Ph.D. program; Professor Joonki Suh; Namwook Hur, student in the integrated M.S.–Ph.D. program; and Dr. Wenxuan Zhu, postdoctoral researcher. >   Building next-generation semiconductors and low-power electronic devices requires precisely stacking materials with different functions. In this process, it is essential to preserve each material's intrinsic properties, as well as the interface where the two materials meet, without damage. Layered van der Waals materials, including transition metal dichalcogenides (TMDs), have attracted considerable attention as next-generation semiconductor platforms because their layers interact through weak forces, enabling different materials to be stacked while maintaining atomically clean interfaces. KAIST (President Choongsik Bae) announced on the 27th of July that a research team led by Professor Joonki Suh from the Department of Chemical and Biomolecular Engineering, in collaboration with Professor Bonggeun Shong's team at Hanyang University and Professor Yimo Han's team at Rice University in the United States, has developed a new semiconductor manufacturing technique based on atomic layer deposition (ALD). ALD is a thin-film deposition process in which semiconductor precursor are supplied sequentially, enabling uniform thin films to be deposited with atomic-level control over their thickness. The research team focused on van der Waals materials. These two-dimensional semiconductor materials consist of multiple atomic layers held together by weak interlayer forces, allowing them to be peeled apart into sheets as thin as paper. Because different materials can be freely stacked, van der Waals materials are attracting attention as key building blocks for next-generation AI chips and ultra-low-power semiconductor devices. However, their chemically stable surfaces make it difficult to grow new semiconductor layers in a uniformly aligned orientation. This challenge becomes even greater at lower temperatures, where atoms tend to nucleate and grow in random directions, making it difficult to produce high-performance semiconductor films. The research team developed a new method that allows tellurium (Te)-containing precursors—molecular building blocks used to fabricate semiconductors—to move freely across the surface, find the most energetically stable positions, and form a thin film. < Figure 1. Diffusion-steered epitaxial atomic layer deposition of tellurium and structural character of tellurium grown on WSe₂. The figure illustrates how precursors first adsorb weakly onto the van der Waals surface, diffuse across it, and initiate tellurium crystal growth at energetically stable positions. Electron microscopy and optical and structural analyses confirmed that the tellurium grew along a consistent crystallographic orientation, forming a clean interface between the two materials while maintaining extremely low interfacial distortion. The results also demonstrate that, unlike conventional high-temperature growth processes, this technique enables epitaxial growth at a low temperature of 150°C. >   Tellurium is attracting attention as a key material for next-generation semiconductors and optoelectronic devices, including photodetectors and light-emitting diodes (LEDs), because it combines highly direction-dependent electrical conductivity with excellent light-controlling properties. Using this approach, the research team succeeded in achieving epitaxial growth of tellurium uniformly in a single direction on van der Waals materials— next-generation two-dimensional semiconductor materials composed of multiple atomically thin layers stacked like sheets of paper— at a low temperature of 150°C using ALD. In epitaxial growth, a new semiconductor film grows in an ordered manner following the atomic arrangement of the underlying crystal, enabling the precise fabrication of high-quality semiconductor films. This process is comparable to stacking bricks neatly in the same direction rather than placing them randomly, which can improve electrical transport and enhance semiconductor performance. The researchers also confirmed that the technology could be applied to a range of van der Waals materials, including tungsten diselenide (WSe2), molybdenum disulfide (MoSS), rhenium diselenide (ReSe2), and mica. This demonstrates that the method is not limited to a single material but can be broadly used with various next-generation semiconductor materials. The team further used the resulting semiconductor films to fabricate transistors, key semiconductor devices that control the flow of electrical current, as well as optoelectronic devices that detect or emit light. This demonstrated that the new manufacturing technology is not confined to laboratory-scale material growth but can also be applied to the fabrication of functional semiconductor devices. < Figure 2. Conceptual illustration of tellurium thin-film growth using diffusion-steered epitaxial atomic layer deposition. The image depicts the process in which two types of precursors are supplied onto a van der Waals surface, diffuse across the surface, and form a tellurium thin film aligned in a consistent crystallographic direction at energetically stable positions. (AI-generated image) >   "This study is the first to demonstrate that high-quality semiconductor films can be grown on van der Waals materials at low temperature without damaging the underlying materials," said Professor Suh. "We expect this technology to serve as a key manufacturing platform for integrating a wide range of next-generation semiconductors on a single chip," he added. The study, with Changhwan Kim, a doctoral student, as first author and Professor Suh as corresponding author, was published in the journal Science Advances on July 24.   Paper title: Van der Waals template-encoded soft epitaxy of tellurium enabled by atomic layer deposition DOI: 10.1126/sciadv.aef1430   Author information: Changhwan Kim (Korea Advanced Institute of Science and Technology / Ulsan National Institute of Science and Technology, first author) and Joonki Suh (Korea Advanced Institute of Science and Technology, corresponding author) This research was supported by the National Research Foundation of Korea, under the Ministry of Science and ICT.  

KAIST Develops Ultra-Precise Inspection Technology..

  < The research team. From left: Jinwoo Jeon, a student in the integrated M.S.–Ph.D. program; Dr. Younggeun Lee; Dr. Guseon Kang; Professor Young-Jin Kim; Jaeyoon Kim, a Ph.D. student; and Dr. Dongwook Yang. >   An ultra-precise inspection technology that could help prevent electric vehicle battery fires and improve battery safety has been developed. A KAIST research team has developed a method capable of detecting minute variations in battery electrode thickness that can contribute to thermal runaway with a precision equivalent to approximately one ten-thousandth the diameter of a human hair, all without disassembling or damaging the battery. The technology is expected to improve battery safety and quality by identifying invisible defects during the manufacturing process. KAIST (President Choongsik Bae) announced on 23rd of July that a research team led by Professor Young-Jin Kim from the Department of Mechanical Engineering has developed a technology that measures the thickness of lithium-ion battery electrodes in a non-contact and non-destructive manner. The technology combines terahertz waves (electromagnetic waves in the spectral region between light and radio waves) to obtain information from inside battery electrodes with an optical frequency comb, which divides the frequency of light into evenly spaced intervals like the markings on a ruler and serves as a reference for ultra-precise measurements. The electrodes in lithium-ion batteries, which are widely used in electric vehicles, are essential components through which electric current flows. Even a slight variation in electrode thickness can cause current to become concentrated in certain areas when charging and discharging, generating heat. If the heat continues to accumulate, it may lead to thermal runaway, a phenomenon in which the internal temperature of a battery rises rapidly and can result in a fire or explosion. Maintaining uniform electrode thickness is therefore critically important during battery manufacturing. Existing inspection technologies, however, have limitations when applied to production environments. X-ray computed tomography can provide detailed images of internal structures, but its relatively long inspection time makes it difficult to use on high-speed production lines. Ultrasonic acoustic microscopy requires direct contact with a liquid medium, while laser displacement sensors can perform rapid measurements but have difficulty precisely analyzing structures inside an electrode. The research team overcame these limitations by combining optical frequency comb and terahertz technologies. The researchers first directed terahertz waves at a battery electrode and collected signals generated as the waves were repeatedly reflected within the electrode. They then used an optical frequency comb as a reference to analyze the signals with exceptionally high precision and calculate the electrode thickness. This enabled nanometer-scale measurements of the electrode’s internal structure without damaging the battery. At the core of the technology is Fabry–Pérot interference, a regularly spaced interference pattern produced as terahertz waves repeatedly travel back and forth between the front and rear surfaces of an electrode. Much like measuring length by reading the markings on a ruler, the researchers precisely analyzed the interference pattern using the optical frequency comb as a reference to determine the electrode thickness. As a result, the team successfully measured both the electrode thickness and its complex refractive index (a material’s optical property indicating how strongly it transmits and absorbs electromagnetic waves) in a single measurement without requiring a separate calibration process. The researchers validated the technology using battery electrodes measuring between 50 and 150 micrometers in thickness, comparable to the diameter of a human hair. With a measurement time of just 0.2 seconds, the system detected thickness differences as small as 70.1 nanometers in the anode (approximately one fourteen-hundredth the diameter of a human hair) and 465.5 nanometers in the cathode. This measurement speed is considered sufficient for use on rapidly moving battery production lines. When the measurement time was increased to 25.6 seconds, the precision improved further. The system distinguished differences as small as 7.8 nanometers in the anode (approximately one ten-thousandth the diameter of a human hair) and 25.2 nanometers in the cathode. This represents up to a 100-fold improvement in precision compared with conventional time-domain analysis methods, enabling the detection of thickness variations that are completely invisible to the naked eye. The technology is not limited to measuring thickness at a single point. It can generate a three-dimensional map of thickness across an entire electrode and track gradual thickness variations in real time during production. The researchers also confirmed that the system could accurately measure an electrode tilted at an angle of approximately 45 degrees, demonstrating its potential for application to fast-moving, real-world battery manufacturing lines.   < 1. Lithium-Ion Battery Electrode Thickness Measurement Using Optical-Frequency-Comb-Referenced Terahertz Fabry–Pérot Interferometry >     The study is significant because it presents a new inspection technology capable of identifying invisible microscopic defects during production without disassembling or damaging batteries. In addition to lithium-ion batteries, the technology is expected to serve as a key quality-control tool for manufacturing next-generation all-solid-state batteries, which use solid electrolytes instead of liquid electrolytes. By detecting defects at an early stage, the technology could improve battery safety and quality while enabling more stable manufacturing processes. “This technology is an integrated metrology platform that can simultaneously measure electrode thickness and material properties without requiring a separate calibration process,” said Professor Kim. “We expect it to become a key technology for the real-time quality control of production lines for next-generation lithium-ion batteries and all-solid-state batteries.” The study was led by Dr. Guseon Kang from the KAIST Department of Mechanical Engineering, currently with the Korea Institute of Industrial Technology, as the first author, with Professor Young-Jin Kim serving as the corresponding author. The research findings were published in the international journal Nature Communications on June 10.   Paper title: Nanometre-precision terahertz interferometry for battery electrode metrology DOI: https://doi.org/10.1038/s41467-026-74193-8   This work was financially supported by the National Research Foundation of Korea (NRF) (RS-2024-00401786, RS-2025-00523273, RS-2025-25455397, RS-2026-25540567, and NRF-2022M1A3C2069728) and from the Korean government’s Defense Acquisition Program Administration (DAPA) (KRIT-CT-22-040).      

KAIST Opens a New Era of Webtoons： From “Viewing” ..

  < Members of the research team. From left: Ammar Al-Taie, a postdoctoral researcher at the KAIST Information and Electronics Research Institute; Professor Ian Oakley of the School of Electrical Engineering; and doctoral student Hyunyoung Han. >   Webtoons are coming to life in the physical world, ushering in a new era in which comics are not merely viewed, but experienced.   A KAIST research team has developed the world’s first next-generation extended reality (XR) comics platform that enables a wide range of readers to enjoy immersive, three-dimensional comics in physical space. By expanding webtoons beyond the screen and into the real world, the team has opened up new possibilities for the future of comics.   KAIST (President Choongsik Bae) announced on the 21st of July that a research team led by Professor Ian Oakley from the School of Electrical Engineering has proposed core design principles and future directions for next-generation extended reality (XR) comics through a systematic user study involving 15 participants, including human-computer interaction (HCI) experts, professional webtoon creators, and readers. < Figure 1: A comic summarizing our study: Comics evolve to meet reader needs, and the digital age has further changed their layouts. To address the next stage in their evolution, we investigated how comics may be laid out in eXtended Reality (XR). >     The research team developed ComiXR, a new platform that enables users to both read and create comics in XR environments. Participants used the platform to transform a conventional print comic into an XR comic and explored how different visual, auditory, haptic, and interactive features could be combined.   Comics, which originated in printed books and newspapers, have evolved dramatically with the rise of smartphones. The vertical-scrolling format of webtoons has become particularly successful by adapting comics to the interaction methods of mobile devices.   The research team viewed XR devices as a potential next stage in this evolution. To explore how spatial depth, three-dimensional rendering, spatial audio, eye tracking, and facial expression tracking could be incorporated into comics, the team built ComiXR using a Meta Quest Pro headset.   < Figure 2: Screenshots of participants using ComiXR. Left: ComiXR’s intitial state; passthrough with menus to spawn XR features. Middle: a participant testing eye-tracking to highlight a character. Right: a participant triggering a speech bubble through eye-tracking by looking at the character. >     While wearing the headset, participants freely positioned 3D characters, speech bubbles, sound effects, and other comic elements throughout a physical room. They were able to construct comic environments that they found comfortable, engaging, and immersive.   The results showed that readers strongly preferred designs that actively used the depth of physical space over simply displaying flat comic pages in a virtual environment. Immersion increased significantly when characters were positioned at a different depth from the background and speech bubbles were separated into distinct layers. In particular, an eye-tracking feature that revealed the next line of dialogue only when the reader looked at a specific character proved effective in preventing spoilers.   The platform also demonstrated new sensory experiences that are not possible in conventional comics. Special effects could be triggered in response to readers’ facial expressions, while haptic feedback could convey sensations such as a character’s heartbeat or the impact represented by an onomatopoeic effect.   < Figure 3. The study setup. First, a 15-minute introduction phase with participants using ComiXR to experience different XR comic features. Second, a 45-minute transformation phase, in which participants adapted a print comic (top left) into XR (top right). Finally, participants answered a post-study survey indicating the acceptability of XR comics. >     Based on the study, the research team also proposed four key design concepts for XR comics. The first, “The Panel Gallery,” transforms the walls of a room into a gallery for displaying comic panels. The second, “The Pop-Up,” presents comics like pop-up books on desks or walls. The third, “Around Comic,” places 3D characters and other comic elements in outdoor spaces. The fourth, “Inclusive ComiX,” improves accessibility for a wide range of readers.   The research team expects XR comics to complement, rather than replace, existing smartphone-based webtoons. They could be used for special exhibitions and educational content that allow audiences to experience fictional worlds more vividly, as well as platforms that improve access to cultural content for a wider range of users.   Ammar Al-Taie, a postdoctoral researcher at the KAIST Information and Electronics Research Institute, participated as the first author, while Hyunyoung Han, a doctoral student in the School of Electrical Engineering, participated as a co-author.   The research was presented at the ACM Designing Interactive Systems Conference 2026, or ACM DIS 2026, one of the leading international conferences in human-computer interaction and design. The ComiXR platform has also been released as open-source software for public use.   Paper title: ComiXR: Exploring Comic Layouts in eXtended Reality DOI: https://doi.org/10.1145/3800645.3812857 Related Video: https://drive.google.com/drive/folders/1D9Efp3T0biDbUm1K5Gu6HLSSy89Uaq8R?usp=sharing Open-source platform: https://github.com/ammarjamal/ComiXR   The research was supported by the KAIST Jang Young Sil Fel¬lowship Program (Excellence Track). The authors acknowledge support from the IITP (Institute of Information & Communications Technology Planning & Evaluation)-ITRC (Information Technology Research Center) grant funded by the Korean government (Ministry of Science and ICT) (IITP-2026-RS-2024-00436398).

KAIST Makes Buttons Rise with Light​

  < From left. Professor Il-Kwon Oh and Hyunsoo Kim, a master’s student in the Department of Mechanical Engineering >   No wires. No actuators. Shine light on the metal surface, and it rises like a button. KAIST researchers have developed a metal structure that changes shape using light, without any light-absorbing coating. This technology could open new possibilities for tactile interfaces with physical pop-up buttons, shape displays, next-generation wearable devices, and soft robots.   KAIST (President Choongsik Bae) announced on the 20th of July that a research team led by Professor Il-Kwon Oh from the Department of Mechanical Engineering has developed a technology that transforms a flat NiTi shape-memory alloy (SMA) sheet into a “photothermally driven meta-morphing structure” that rises from a flat surface into a three-dimensional form when exposed to light, using only a single UV-laser process.   Next-generation wearable devices and soft robots require technologies that are thin and lightweight while also being capable of changing into desired shapes when needed. Such technologies are attracting attention as a foundation for shape displays, adaptive surfaces, wearable interfaces, and soft robots. < Figure 1. Concept and monolithic fabrication strategy of the UV laser-programmed photothermal SMA kirigami platform. >     The research team designed precise cutting and folding patterns in a flat metal sheet so that it would transform into a predetermined three-dimensional shape. The design principle is based on kirigami, the art of creating three-dimensional structures by cutting paper.   Shape-memory alloys are special metals that return to a pre-programmed shape when heated to a specific temperature, even after being deformed. Because they are lightweight and can generate large forces, they are widely used as key materials for soft robots and wearable actuators. However, conventional photothermal shape-memory alloy actuators have faced a limitation: nickel-titanium alloy (NiTi) surfaces do not absorb near-infrared light efficiently. To compensate for this, separate light-absorbing coatings such as graphene oxide, polymer composites, or titanium nitride (TiN) thin films have typically been applied.   These external coatings can peel off during repeated operation and require additional processing. They can also increase heat capacity, which may slow the response, creating limitations in both manufacturability and actuation performance.   To address this problem, the research team used UV laser micromachining. Through this process, they formed kirigami structures on thin shape-memory alloy (SMA) sheets while simultaneously generating a micro-nano porous titanium oxide (TiOₓ) layer on the surface through laser-induced oxidation. As a result, they were able to significantly increase the absorption of near-infrared light without any separate external coating.   The team also implemented a platform that can precisely control the height of three-dimensional deformation and the resulting force output by adjusting structural parameters such as hinge width and slit width. In other words, the core of this research lies in simultaneously programming both how the structure mechanically deforms and how efficiently it absorbs light within a single metal structure.   Furthermore, the team applied a patterning technique that spatially controls the degree of laser-induced oxidation. This made it possible for different regions to deform sequentially at different speeds, even when exposed uniformly to light of the same intensity. The researchers describe this as “spatiotemporal actuation control.” This means that the order and timing of deformation are encoded directly into the material itself through light-absorption properties, without any separate electrical control. It can be seen as a form of photonic logic.   The research team further expanded the photothermal SMA metastructure into three-dimensional shape displays and haptic interfaces by integrating it with a multi-channel near-infrared (NIR) LED array. Each SMA kirigami unit moves independently in response to selectively applied light. Based on this, the team successfully displayed the letter sequence K→A→I→S→T and implemented tactile navigation signals that indicate direction. < Figure 2. Demonstration of three-dimensional letter patterns (K→A→I→S→T) and directional navigation haptic cues, achieved by selectively illuminating and independently actuating individual units of the shape-memory alloy meta-morphing structure. >   Professor Il-Kwon Oh said, “The laser programming technology developed in this study is a manufacturing-friendly platform that encodes both mechanical deformation and optical properties into a single metallic structure without any separate coating process,” adding, “It can be widely applied to next-generation intelligent morphing interfaces controlled by light, including adaptive surfaces, interactive haptics, and photothermal soft robots.”   Hyunsoo Kim, a master’s student in the Department of Mechanical Engineering, served as the first author, while Professor Il-Kwon Oh was the corresponding author. The results were published in the international journal Advanced Science, and the work was also selected for the Inside Back Cover of Advanced Science, Vol. 13, No. 31, published on June 4, 2026.   < Figure 3. Image of the photothermal shape-memory alloy meta-morphing structure featured on the Inside Back Cover of Advanced Science >     Paper title: Monolithic UV-Laser Programming of Photothermally Meta-Morphing SMA Structures: Dual-Encoded Kirigami Mechanics and Photonic Absorbance DOI: https://doi.org/10.1002/advs.74930 Related Video: https://drive.google.com/drive/folders/16Q9C1D6EMjoM2ypNmtDGU9ruMBdZeUFs   This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (RS-2024-00345241 and RS-2023-00302525). This research was supported by the Nano & Material Technology Development Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Science and ICT (RS-2025-25441263). This research was supported by the InnoCORE program of the Ministry of Science and ICT (N10250154).      

KAIST’s Solarstill Box Wins Red Dot’s Highest Hono..

  < Professor Bae Sang-min (in a blue vest), his research team, and local residents during a field study and co-design workshop in Tanzania. >     A KAIST design that produces clean drinking water using only sunlight, without electricity or fuel, has won one of the world’s most prestigious design awards. The design received high international recognition not only for its technical completeness, but also for its sustainable approach, which enables residents in regions affected by water scarcity and water pollution to produce and manage clean water on their own.   < Figure 1. Highlighting Local Drinking Water Challenges and the Need for the SolarStill Box >     KAIST (President Choongsik Bae) announced on the 17th of July that “Solarstill Box,” a solar-powered water purification and desalination device developed by a research team led by Professor Sangmin Bae from the Department of Industrial Design, has won the “Red Dot: Best of the Best” award in the Social Impact category at the Red Dot Award: Design Concept 2026, a globally renowned design competition.   The Red Dot Design Award is considered one of the world’s three major design awards, along with Germany’s iF Design Award and the United States’ IDEA (International Design Excellence Awards). Among them, the “Best of the Best” is the highest distinction, awarded to works that demonstrate the greatest innovation and completeness in each category.   < Figure 2. The SolarStill Box prototype in operation >     This award is significant because it goes beyond recognition of product design excellence. It represents international acknowledgment that design can help address the shared human challenge of clean water access and create sustainable social value.   Solarstill Box is a low-cost water purification and desalination device that converts seawater or water containing salt and pollutants into drinking water through solar distillation. It was developed for coastal areas, saline regions, and off-grid communities that rely on contaminated water sources. The device is designed to produce clean water using only solar energy, without electricity, fuel, or separate filters. < Figure 3. Structure of the solar-powered water purification and desalination system >     The stepped trays inside the device increase the surface area for evaporation, improving the efficiency of solar distillation. As evaporated water vapor condenses on the transparent cover, contaminants such as salt, heavy metals, and bacteria are naturally separated, allowing only clean water to be collected.   < Figure4. Conceptual diagram showing how solar heat evaporates contaminated water, allowing only clean water to be separated and collected >   Solarstill Box is also made of flat components based on Plaveneer sheets, allowing it to be produced and transported in a flat-pack format. Anyone can assemble it locally in about 20 minutes, and maintenance costs are reduced because only damaged parts need to be replaced. Another key feature is that it moves beyond the one-time delivery of relief supplies and instead creates a sustainable drinking water system that local residents can install and manage themselves.   Solarstill Box was developed as part of the “SEED Project,” a social contribution design research project by ID＋IM Design Lab, led by Professor Sangmin Bae of the Department of Industrial Design, in collaboration with World Vision. Following this award, the research team plans to work with World Vision to pursue product commercialization and establish local distribution models. In the long term, the team aims to expand the project into a cooperative-based operating model in which local residents directly participate in manufacturing, distribution, and maintenance, thereby supporting both clean water access and sustainable community development.   < Figure 5. A sustainable donation-based model for local assembly, operation, and distribution of the SolarStill Box >     The project aimed to improve access to clean drinking water in low-resource regions where water scarcity and pollution make it difficult to secure safe water. The design development was carried out by Professor Sangmin Bae, doctoral student Jungwoo Kim, master’s student Minsu Kim, and undergraduate student Seunghee Han.   Professor Sangmin Bae said, “Design should go beyond creating beautiful products; it should serve as a tool for solving social problems and changing people’s lives,” adding, “We hope Solarstill Box will provide practical help to communities in need of clean water and spread as a sustainable drinking water system that residents can operate on their own.”   Solarstill Box will be introduced to a global audience through the official award ceremony and exhibition of the Red Dot Award: Design Concept 2026, which will be held in October.        

KAIST Alumnus and TeamSparta CEO Beomgyu Lee Donat..

  < (From left) Professor Sukyoung Ryu from the KAIST School of Computing; Beomgyu Lee, CEO of TeamSparta; Jae-Gil Lee, Dean of the KAIST School of Computing >   KAIST (President Choongsik Bae) announced on the 16th of July that Beomgyu Lee, CEO of AI education company TeamSparta, donated 100 million KRW (approximately ＄66,000 as of July 2026) to the School of Computing to support a research program centered on the use of generative AI agents.   With the support of this donation, KAIST plans to run a two-year program enabling master’s and doctoral researchers from all departments to use generative AI agents. Participating researchers will receive access to the latest AI development agents, including Claude Code and Codex, along with related training and regular seminars.   The donation ceremony was held on the 15th of July at KAIST's main campus in Daejeon and was attended by Beomgyu Lee; Jae-Gil Lee, Dean of the KAIST School of Computing; and Professor Sukyoung Ryu from the School of Computing. This marks Lee’s second private contribution, following his first donation in September 2025.   The program will operate in six-month cycles, with 20 researchers selected for each cohort. KAIST aims to select the first cohort by the end of 2026 and plans to gradually expand both the scope of support and donor participation.   The program is meaningful in that, amid the rapid spread of generative AI across research settings, it lays the groundwork for KAIST researchers to be early adopters of cutting-edge AI tools and apply them to their research. In particular, the program is expected to foster a new research culture in which master’s and doctoral researchers actively incorporate AI agents into their research workflows and share their experiences with one another.   “I sincerely thank alumnus Beomgyu Lee for once again making a meaningful contribution in support of junior researchers,” said KAIST President Choongsik Bae. “His generous commitment will provide a strong foundation for researchers to actively use generative AI and pursue new research challenges. KAIST will continue to provide sustained support so that researchers can produce creative and innovative outcomes in the best possible environment during this era of AI transformation.”   “AI agents have already become a new standard across industry, and I was concerned that researchers who will lead the development of future technologies were unable to make full use of them because of the associated costs,” said Beomgyu Lee. “I hope this support will help researchers freely use cutting-edge AI tools to help produce world-class research outcomes.”   Professor Sukyoung Ryu from the KAIST School of Computing said, "I am deeply grateful to alumnus Beomgyu Lee for continuing to support junior researchers for a second year in a row." She added, “As generative AI agents are significantly enhancing research productivity and transforming approaches to problem-solving, I hope this program will encourage researchers to actively use the latest AI tools and foster a new culture of sharing their experiences.”   Jae-Gil Lee, Dean of the KAIST School of Computing, said, “We will actively support training and regular seminars so that researchers across all KAIST departments—not just the School of Computing—can use the latest generative AI agents in their research, allowing AI-driven research innovation to spread across the university.”  

Professor Seung-Young Ahn Receives Minister of Lan..

  <Professor Seung-Young Ahn>   Professor Seung-Young Ahn of our university has received a commendation from the Minister of Land, Infrastructure and Transport in recognition of his contributions to establishing an advanced talent development system for the railway sector and strengthening the competitiveness of Korea’s railway industry.   In August 2024, Professor Ahn planned and established the KAIST-KORAIL Railway Mobility Program, which selects outstanding employees of Korea Railroad Corporation, or KORAIL, and supports them in pursuing master’s degrees at the Korea Advanced Institute of Science and Technology, or KAIST. The program is the first contract-based academic program established by KORAIL in partnership with an external university. It was created to systematically cultivate professionals capable of leading technological innovation in the future railway sector.   Beginning in the spring semester of 2025, the KAIST-KORAIL Railway Mobility Program will select approximately 20 highly qualified participants each year for five years. The program supports advanced coursework, cutting-edge technology research, and international technical exchanges. Through this initiative, an organized talent development framework has been established to help railway professionals with field experience acquire the expertise and research capabilities required by the future railway industry.   The program has also introduced full-time education and intensive coursework during academic breaks to improve learning efficiency. In addition, a comprehensive administrative and financial support system covers all tuition and educational expenses, enabling participants to focus fully on their studies and research.   Professor Ahn played a central role in planning and establishing the program, designing its curriculum and operating structure, and building the foundation for industry-academia cooperation between KORAIL and KAIST. His efforts have been recognized for cultivating key professionals who can advance railway systems and identify new business opportunities, while also laying the groundwork for research into advanced railway technologies.   The commendation recognizes Professor Ahn’s contribution to transforming KORAIL’s approach to talent development through the establishment of the corporation’s first external contract-based academic program and the creation of a professional workforce development system designed to lead future railway innovation.   Professor Ahn said, “I will work to ensure that the KAIST-KORAIL Railway Mobility Program becomes a leading industry-academia cooperation platform that addresses technical challenges in the railway field and cultivates key professionals who will lead the future railway industry.” He added, “I will continue to contribute to strengthening the competitiveness of Korea’s railway industry by promoting the convergence of advanced mobility technologies and the railway sector.”    

KAIST Develops Robot That Judges Its Surroundings ..

  < People. The research team. From left: Ph.D. candidate Jaehyun Park (KAIST, co-first author); Professor Hae-Won Park (KAIST, corresponding author); Professor Seungwoo Hong (Korea University, corresponding author); Researcher Jun-Gill Kang (Agency for Defense Development at the time of the research, co-first author). >   An era in which robots decide "how to walk" on their own has arrived. A four-legged robot has been developed that, much like a person or an animal, autonomously chooses the appropriate gait strategy for its surroundings — changing its gait on stairs, leaping over gaps, and keeping its balance on forest trails.   KAIST (President Choongsik Bae) announced on the 16th of July that a research team led by Professor Hae-Won Park from the Department of Mechanical Engineering has developed a core control technology for four-legged robots that lets a single controller select and switch in real time among walking, running, jumping, and other locomotion skills, allowing the robot to move quickly and stably, even in real outdoor environments. < Figure 1. KAIST HOUND demonstrates its ability to overcome various obstacles using the newly developed control technology. >   Four-legged robots move on four legs, giving them an advantage over wheeled robots on rough terrain. But in real outdoor settings, obstacles such as stairs, ledges, stepping stones, gaps, and tree branches appear one after another in different forms, meaning the ability to simply walk and run fast is not enough. Existing four-legged robots have excelled at running quickly across flat ground or clearing simple obstacles, but they have struggled to maintain both speed and stability in real-world environments where obstacles combine in complex ways. Because walking, running, jumping, and other gaits had to be controlled individually, the robots were also limited in how naturally they could switch between them as conditions changed.   To overcome these limitations, the research team developed a new learning-based control technology called APT-RL (Action Pretrained Transformer-based Reinforcement Learning).   APT-RL is a control technology designed to enable a robot to first learn a range of locomotion skills — such as walking, running, and jumping — and then freely combine and transition among them in real-world environments as the situation demands.   Rather than filming the movements of real people or animals, the team generated 15.5 hours of training data covering a variety of gaits using computer simulations alone, in just eight minutes. That data was used to teach the robot basic movement capabilities, drawing on robot dynamics (a mathematical model of how a robot moves) and trajectory optimization (a technique for calculating the efficient path of movement). The approach is far faster and more efficient than earlier methods that relied on motion capture, a technology that records human or animal movement using sensors. < Figure 2. Overview of the developed control technology >     The team then applied reinforcement learning — an artificial intelligence technique in which an agent learns optimal behavior through repeated trial and error — so the robot could autonomously select and switch gaits suited to complex three-dimensional terrain such as stairs, ledges, and gaps. Finally, the team combined a depth camera (which measures the distance to objects in order to obtain three-dimensional information) with LiDAR (Laser Detection and Ranging, a sensor that uses lasers to measure the distance and shape of the surrounding environment in three dimensions), enabling the robot to recognize its surroundings and target speed in real time and choose the most appropriate walking strategy.   The team tested the control technology on its own four-legged robot, 'KAIST HOUND.' The experiments were conducted not only on an indoor obstacle course but also in real outdoor environments, including KAIST’s campus and forest trails.   KAIST HOUND moved stably across urban terrain that included stairs, grass, and slopes, as well as irregular natural terrain such as fallen trees, exposed roots, and paths covered in fallen leaves, switching gaits in real time to match the conditions. In rugged terrain with obstacles, the robot reached a peak instantaneous speed of six meters per second (about 22 kilometers per hour), demonstrating that it can achieve both fast movement and stability in real outdoor environments.   The experiments showed that KAIST HOUND autonomously selected and switched between a trot (alternating diagonal legs) and a bound (a leaping gait using the front and back leg pairs together) depending on the terrain and target speed, and that it could integrate walking, running, jumping, and ledge-clearing into a single controller.   Professor Hae-Won Park said "We expect this to become a foundational technology that expands the potential uses of physical-AI-based walking robots in rugged environments such as disaster sites, defense missions, and industrial facility inspections."   Jun-Gill Kang (affiliated with the Agency for Defense Development (ADD) at the time of the research) and Jaehyun Park, a Ph.D. candidate in KAIST's Department of Mechanical Engineering, are co-first authors of the study. Professor Hae-Won Park and Professor Seungwoo Hong from Korea University are co-corresponding authors. The research was selected as the cover paper for the July issue of Science Robotics, the world's leading academic journal in robotics, and was published on July 15 (U.S. Eastern time). < Cover image of the July issue of Science Robotics >   Paper title: Agile perceptive multi-skill locomotion for quadrupedal robots in the wild DOI: 10.1126/scirobotics.adz7397 Authors: Jun-Gill Kang (the Agency for Defense Development at the time of the research, co-first author), Jaehyun Park (KAIST, co-first author), Hae-Won Park (KAIST, corresponding author), Seungwoo Hong (Korea University, corresponding author)   This research was supported by funding from the Ministry of Trade, Industry and Resources (MOTIR) and the Korea Planning & Evaluation of Industrial Technology (KEIT) (RS-2024-00427719), as well as by the Agency for Defense Development's Future Challenge Defense Technology R&D program (912768601).  

KAIST Develops Key Technology to Make Personalized..

  < The research team. From left: Ph.D. candidate Wonjun Lee, Professor Changick Kim, Ph.D. candidate Seokil Ham, and Ph.D. candidate Jaehyuk Jang. >   “Create an AI assistant trained only on our company’s documents.”   The era of building “personalized AI” by training AI models on individual or corporate documents and data is beginning. However, while such customization can improve task performance, it can also weaken the model’s existing safety safeguards. KAIST researchers have developed a core AI technology that preserves customized performance while further strengthening safety.   KAIST (President Choongsik Bae) announced on the 15th of July that a research team led by Professor Changick Kim from its School of Electrical Engineering has developed “Buffer-and-Reinforce,” a training framework for safe fine-tuning that prevents safety degradation when large language models (LLMs), such as    ChatGPT, are retrained on data from individuals or companies to better suit their needs.   Until now, one of the biggest challenges in the era of personalized AI has been that fine-tuning improves a model’s ability to perform new tasks, but can also weaken its existing safety rules. The research team focused on prior findings showing that, counterintuitively, fine-tuning an AI model while it is in a temporarily jailbroken state — a state in which it may respond even to dangerous requests it would normally refuse — does not significantly compromise its safety.   The team then devised a new approach in which this jailbroken state is not used in actual services, but is applied only temporarily during the fine-tuning process through a buffering module called “BufferLoRA,” which is removed after training.   The research team was the first to clarify why this phenomenon occurs. They found that, in the temporarily jailbroken state, the AI model becomes less easily influenced by harmful information, while still effectively learning the new task abilities desired by the user. In other words, the model can continue learning useful knowledge without additionally absorbing harmful behaviors.   Based on this insight, the team developed a two-stage learning method consisting of “buffering” and “safety reinforcement.”   First, the temporary buffering module, BufferLoRA, is applied to the AI model during user fine-tuning, where it acts as a protective layer that prevents harmful data from directly affecting the base model. Once fine-tuning is complete, this module is removed. Next, a safety reinforcement module called “ReinforceLoRA” is applied to restore and strengthen the model’s safety. In this process, the team used QR decomposition, a mathematical technique that separates different types of information and selectively reflects only the necessary components. This allowed the model to retain the new functions learned from user data while selectively reinforcing safety.   Simply put, the researchers first placed a temporary protective layer, BufferLoRA, over the AI model so that harmful data could not directly affect it, while allowing the model to learn the necessary task. They then removed the protective layer and applied ReinforceLoRA to strengthen the model’s safety safeguards. As a result, the model maintained its customized performance while achieving even stronger safety. < Figure 1. Infographic of the Buffer-and-Reinforce training framework and its applications. >     In experiments, the AI model maintained high safety even in an extreme setting where all user data consisted of harmful questions and answers. After fine-tuning, the rate at which the AI generated harmful responses was about 8％, lower than the roughly 18％ observed in the original model that had not been fine-tuned at all. The framework also achieved strong customized performance and state-of-the-art safety without requiring additional safety data during user fine-tuning or significantly increasing computational cost, suggesting its practical applicability to real-world personalized AI services.   Professor Changick Kim stated, “This research provides a key foundational technology that allows anyone to build customized AI with their own data while using it more safely,” adding, “We expect it to contribute significantly to building a trustworthy AI service environment in the era of personalized AI and AI agents.”   This research was led by Seokil Ham, a doctoral student in KAIST’s School of Electrical Engineering, as first author. The paper was selected as a Spotlight presentation at the International Conference on Machine Learning (ICML) 2026, one of the world’s most prestigious conferences in artificial intelligence, an honor given to only about the top 2.2％ of all submitted papers, drawing international attention.   ※ Paper title: Jailbreak to Protect: Buffering and Reinforcing via Temporary Jailbreaking for Safe Fine-Tuning in Large Language Models DOI: 10.48550/arXiv.2605.24550 ※ Author information: Seokil Ham (KAIST, first author), Jaehyuk Jang (KAIST, second author), Wonjun Lee (KAIST, third author), Changick Kim (KAIST, corresponding author) ※ Related video: https://drive.google.com/file/d/1gfok06dE8699qtiUR7gVsRoVmBGADaWQ/view?usp=sharing   This work was supported by Institute of Information & Communication Technology Planning & Evaluation (IITP) grant funded by Ministry of Science and ICT(MSIT) (No. RS-2025-02215344, Development of AI Technology with Robust and Flexible Resilience Against Risk Factors).