Abstract Robust reinforcement learning (RRL) aims to tackle unexpected environmental changes by optimizing policies against the worst case. However, RRL remains impractical due to the cost of the Max-Min optimization, where it suffers from the exhaustive query complexity for finding the worst-case (dubbed 'Min') within the environmental uncertainty set U , ie, O (|U|). By viewing this via a lens of quantum perspective, we raise a pivotal question: If we can query from the environment with quantum superpositions, is it possible to accelerate the Max-Min optimization of RRL? Our answer is 'Yes'. Our method, called quantum robust inner minimization (QRIM), encodes the uncertainty set with quantum superposition and amplifies low-return cases, thus enabling RL for solving the robust (ie, worst-case) Bellman equation. Importantly, QRIM achieves a quadratic speed-up in query complexity without altering the outer RL pipeline, ie, O (|U|). Validated through classical simulations to real quantum hardware execution, QRIM learns more robust policies with quadratically reduced queries than classical RL. Artificial intelligence systems deployed in the real world must continue to make reliable decisions even when conditions differ from those encountered during training. Robust reinforcement learning (RRL) addresses this challenge by preparing AI agents for the worst-case scenarios they may face. However, identifying those worst-case conditions requires searching a large uncertainty space, making training increasingly expensive as the number of possible environmental variations grows. To this bottleneck, Professor Sung Whan Yoon of the Graduate School of Artificial Intelligence at UNIST, in collaboration with Professor Joongheon Kim of Korea University, developed Quantum Robust Inner Minimization (QRIM), a method that replaces only the inner optimization responsible for identifying worst-case environments, while leaving the overall reinforcement learning pipeline unchanged. QRIM exploits quantum superposition to evaluate many possible environmental conditions simultaneously rather than examining each one individually. By amplifying the worst-case scenarios through a quantum search routine, the method reduces the number of required environment queries from N to √N—a quadratic improvement in query complexity—without modifying the surrounding reinforcement learning procedure. Across a series of experiments, QRIM learned more robust policies while requiring only 20–30% of the computation used by conventional methods. The researchers validated the algorithm through both classical simulations and experiments on IBM's 127-qubit quantum processor, demonstrating that the quadratic speed-up remained effective even in the presence of hardware noise. “QRIM redesigns the most computationally demanding component of RRL—the search for worst-case environments—using a quantum algorithm,” said first author Hyun Kyu Lee. "Because only the inner optimization is replaced, existing reinforcement learning algorithms can be readily incorporate the quantum module, making the approach broadly applicable across robust reinforcement learning." “This study demonstrates how quantum computing can complement artificial intelligence by addressing a well-defined computational bottleneck,” said Professor Yoon. "As quantum hardware continues to advance, approaches like QRIM could help make RRL practical for applications that require reliable decision-making under uncertainty, including robotics and autonomous driving." The findings have been accepted for presentation at the International Conference on Machine Learning (ICML) 2026, one of the world's leading conferences on machine learning. Among the quantum AI papers accepted to ICML 2026, QRIM is the only study led by a Korean research institution. The research was supported by the National Research Foundation of Korea (NRF) and the Institute for Information & Communications Technology Planning & Evaluation (IITP) through programs funded by the Ministry of Science and ICT (MSIT). Journal Reference Hyun Kyu Lee, Joongheon Kim, and Sung Whan Yoon, “Quantum Robust Inner Minimization for Reinforcement Learning with Quadratic Speed-Up in Query Complexity,” ICML, (2026) .

Abstract The explosive growth of open-source model repositories has created a Model Jungle, where checkpoints are frequently shared without adequate documentation or metadata. While weight-space learning offers a pathway to identify and analyze these models directly from their parameters, processing full-scale weights is computationally prohibitive. Probing-based methods have emerged as a lightweight alternative, extracting permutation-equivariant representations via learnable probe vectors. However, existing probing methods are limited by a single-view design: they capture first-order structures but fail to encode the rich, higher-order correlation patterns inherent in row-column interactions. To bridge this gap, we introduce MVProbe, a multi-perspective probing framework that synthesizes first-order signals with interaction-aware (Gram-based) views. Our approach is theoretically grounded; We analyze the scaling laws of different probing orders to derive a principled standardization and fusion strategy that ensures balanced contributions from all branches. On the Model Jungle benchmark, MVProbe consistently outperforms the state-of-the-art ProbeX across diverse architectures, including discriminative backbones (ResNet, SupViT, MAE, DINO) and large-scale generative LoRA adapters (Stable Diffusion LoRA). Not every AI model comes with a manual. As open-source model repositories continue to grow, many checkpoints are shared with little information about how they were trained or what data they learned from. As a result, identifying suitable models for reuse—or tracing where they came from—is becoming increasingly difficult. To address this challenge, Professor Jaejun Yoo and his team at the UNIST Graduate School of Artificial Intelligence developed MVProbe (Multi-view Probe), a method that identifies what an AI model has learned by analyzing its internal parameters alone, without relying on metadata or executing the model itself. Conventional probing methods analyze model weights from only a single perspective, limiting their ability to distinguish between closely related models. MVProbe overcomes this limitation by examining neural network parameters from four complementary perspectives, capturing not only individual parameter values but also the relationships among them to build a more complete representation of a model. The researchers also showed why existing probing methods struggle to distinguish similar models, demonstrating mathematically that important information embedded in model parameters is overlooked when weights are analyzed from only a single perspective. By integrating multiple complementary views, MVProbe overcomes this limitation while remaining reliable regardless of which neural network layer is analyzed. The team evaluated MVProbe on the Model Jungle benchmark, a large collection of modern AI models spanning image recognition networks and large generative models. Across all evaluated architectures, the method consistently outperformed the previous state-of-the-art approach, improving identification accuracy by as much as 5.96 percentage points while maintaining stable performance across different network layers. Beyond making large model repositories easier to search and organize, MVProbe could support a wide range of applications, including model reuse, provenance analysis, copyright verification, and AI security auditing—all without requiring models to be executed. “As more pretrained AI models become publicly available, the ability to identify and reuse the right one is becoming increasingly important,” said Professor Yoo. "MVProbe makes it possible to search and manage large model repositories by reading the information already encoded in a model's parameters, without running the model itself." The study was co-first authored by Eunwoo Heo and Kyungguk Seo of the UNIST Graduate School of Artificial Intelligence. It has been accepted for presentation at the International Conference on Machine Learning (ICML) 2026. This work was supported by the Institute for Information & Communications Technology Planning & Evaluation (IITP) and the National Research Foundation of Korea (NRF). Journal Reference Eunwoo Heo, Kyeongkook Seo, and Jaejun Yoo, "What Linear Probes Miss: Multi-View Probing for Weight-Space Learning,"

Abstract Direct photopatterning with photoactivated cross-linkers provides a powerful route to highly integrated, solution-processed electronics. We introduce Diazo-6Bx, a six-branched diazo-based cross-linker that combines high miscibility with organic semiconductors and efficient C–H insertion reactivity. Comparative studies with azide-functionalized cross-linkers (phenyl- and tetrafluorophenyl-based structural analogues, namely, Ph-6Bx and 6Bx, respectively) reveal that the fluorine-free structure of Diazo-6Bx and its superior compatibility with host materials enable highly uniform cross-linking. This results in high-fidelity photopatterning with steep taper angles, excellent pattern retention, and minimal line-width variation. In solution-processed organic thin-film transistors (OTFTs), Diazo-6Bx enables the fabrication of unit devices that outperform those based on 6Bx, as well as large-area, high-resolution transistor arrays and logic circuits with excellent operational stability and scalability. These results demonstrate that a careful cross-linker design will play a crucial role in patterning next-generation electronic materials in extremely high fidelity. Printing electronic circuits from solution offers a promising route to lightweight, flexible electronics, but producing fine circuit patterns without sacrificing device performance remains a persistent manufacturing challenge. Professor BongSoo Kim and his research team in the Department of Chemistry at UNIST, working with collaborators at Yonsei University, have developed a new photo-crosslinker that enables more precise patterning of organic semiconductor circuits while improving the stability of the resulting devices. The material, Diazo-6Bx, was designed to blend more uniformly with hydrocarbon-based organic semiconductors than existing photo-crosslinkers. During ultraviolet patterning, it forms cross-linked networks that preserve circuit geometry, producing cleaner edges and more consistent feature sizes. In photopatterning experiments, Diazo-6Bx reduced line-width variation from 10.3 μm to 2.5 μm—roughly one quarter that of a widely used conventional photo-crosslinker. Circuit edges also became significantly sharper, with taper angles increasing from 67.9° to 87°, approaching nearly vertical sidewalls. The improved material compatibility also reduced charge-trapping defects, resulting in more stable transistor performance during prolonged operation. Organic thin-film transistors fabricated using Diazo-6Bx maintained more stable electrical characteristics under prolonged operation than devices produced with conventional crosslinking materials. The researchers also demonstrated that both p-type and n-type organic semiconductors could be patterned sequentially using the same solvent process. Using the same process, the researchers fabricated an 84-transistor organic thin-film transistor (OTFT) array together with complementary NOT, NAND, and NOR logic gates, demonstrating the scalability of the approach for complex organic electronic circuits. “This photo-crosslinker improves both patterning precision and device stability,” said Professor BongSoo Kim of the Department of Chemistry at UNIST. "Its compatibility with both p-type and n-type organic semiconductors simplifies solution processing and could help simplify manufacturing of large-area flexible displays and wearable electronic devices." The study was conducted in collaboration with Professor Jeong Ho Cho of Yonsei University. The findings were published in ACS Nano on June 23, 2026. The research was supported by the Samsung Future Technology Incubation Program and the National Research Foundation of Korea (NRF) through funding from the Ministry of Science and ICT (MSIT). Journal Reference Hyobin Ham, In Cheol Kwak, Daeyeon Lee, et al. , “Diazo-6Bx, a Six-Branched Diazo Cross-Linker, Enables High-Fidelity Patterning for Solution-Processed Electronics with Stable Operation,” ACS Nano, (2026).

Abstract A proposed chest patch device and its interface IC are configured to allow detection of multi-domain signals in the form of optical, electrical, acoustic, and chemical signals at a single body spot. The IC integrates on-chip classification capabilities for cardiovascular diseases (CVDs) based on photo-plethysmogram (PPG) and electrocardiogram (ECG), along with hazardous gas analysis. For low-power multimodal sensing, an ECG R-peak_triggered PPG window (RPT-PW) is proposed as an inter-sensor scheme to reduce activity on the PPG channel, which is the dominant energy consumer. Additionally, the PPG and ECG readout channels incorporate analog peak detection to enable normalized on-chip extraction of pulse arrival time (PAT) and RR interval (RRI). For minimal computation and communication power, a multi-domain convolutional neural network (MD-CNN) processes both analog and digital inputs, and a reconfigurable analog ternary/binarized neural network (a-TNN/BNN) provides additional computation capability for multi-domain applications. The IC was fabricated in an 180-nm BCD process and integrated into a chest patch device prototype with an in-house adhesive meta patch. The RPT-PW demonstrated adaptive operation with an effective LED duty cycle (EDC) as low as 0.02%. The on-chip computation achieved average sensitivity and specificity of 90.87%/95.38% for three-label hypertension, 92.80%/96.36% for three-label arrhythmia, and 92.46%/97.50% for four-label gas mixture classification. Continuous wearable monitoring often requires transmitting large volumes of sensor data to external devices for analysis, increasing power consumption and limiting operating time. A joint research team, led by Professor Jae Joon Kim of Electrical Engineering and Professor Hoon Eui Jeong of Mechanical Engineering, has developed a wearable chest patch that analyzes physiological and environmental signals directly on the device, reducing both communication and sensing power while enabling continuous real-time monitoring. The wearable integrates optical, electrical, acoustic, and chemical sensing into a single device capable of simultaneously monitoring cardiovascular signals and hazardous gases. Rather than transmitting raw sensor data, it processes the information using an on-chip AI processor and sends only the analysis results via Bluetooth. This reduces communication overhead while allowing multiple users to be monitored remotely. To extend battery life, the researchers introduced an adaptive sensing scheme that selectively activates the optical sensor—the most power-intensive component of the device. By synchronizing optical measurements with ECG signals, the system reduced power consumption in the optical sensing channel by approximately 83%, enabling longer operation on a single charge. The device achieved more than 90% sensitivity and specificity in classifying hypertension and cardiac arrhythmias. It also classified hazardous gas mixtures with 92.46% accuracy, demonstrating its ability to monitor both physiological conditions and environmental hazards using a single wearable platform. The chest patch incorporates a microstructured adhesive interface that conforms securely to rough skin while allowing clean removal without leaving residue, making it better suited for long-term wear. “This work brings multimodal sensing and AI inference together on a single wearable platform,” said Professor Kim. “By processing information directly on the device, the system reduces communication demands while enabling continuous monitoring of both health conditions and environmental hazards.” The study has been participated by Sanghyeon Cho and Hyunjoong Kim as co-first authors. Their findings will appear in the July 2026 issue of the IEEE Journal of Solid-State Circuits (JSSC) , one of the leading journals in integrated circuit design. The technology has been licensed to Anvix Lab, a startup co-founded by the research team, which is pursuing commercialization of next-generation bioelectronic patch platform based on on-chip AI. The research was supported by the Ministry of Trade, Industry and Energy (MOTIE), the Ministry of Education (ME), and the Ministry of Science and ICT (MSIT). Journal Reference Sanghyeon Cho, Hyunjoong Kim, Dong Kwan Kang, et. al., “An Energy-Efficient Chest Patch Interface With Inter-Sensor PPG Windowing and Multi-Domain On-Chip Analog Computing,” JSSC , (2026).

Abstract A surge of recent advancements has consistently highlighted the superiority of multimodal learning over unimodal approaches across a variety of tasks. However, the theoretical foundations elucidating this advantage remain underexplored: existing theoretical analyzes are often constrained by tight assumptions, and lack empirical validation. In this paper, we bridge this gap by proposing a novel theoretical framework grounded in convolutional smoothing, offering a new perspective on how multimodal learning contributes to a smoother loss landscape compared to unimodal learning. Building upon this theoretical foundation, we introduce a simple yet effective distributional training strategy based on stochastic modality pairing instead of a fixed pairing; Thus, further promoting a flatter landscape via convolutional smoothing. Our empirical results across various multimodal datasets demonstrate that multimodal models not only achieve higher performance but also exhibit flatter loss landscape, which represent better generalization and robustness. Artificial intelligence models trained on combinations of images, speech, and text consistently outperform those that learn from a single type of data. While this advantage has been widely observed, the reason multimodal learning produces more accurate and reliable models has remained largely unexplained. Researchers at UNIST have now provided a mathematical explanation for this phenomenon and used it to develop a new training strategy that further improves multimodal learning. The study, led by Professor Sung Whan Yoon of the Graduate School of Artificial Intelligence, shows that learning from multiple data modalities produces a flatter loss landscape—a property associated with models that generalize more effectively to unfamiliar data and remain more resilient to noise and perturbations. The researchers demonstrate that this effect arises through a mathematical mechanism known as convolutional smoothing. By learning from different types of information simultaneously, multimodal models naturally smooth abrupt changes in the optimization process, creating a more stable learning landscape than models trained on a single modality. Building on this theoretical framework, the team developed Distributional Multimodal Learning (DML), a simple training strategy that replaces fixed image-text or image-audio pairs with randomly sampled combinations drawn from the same class. This broader variety of training examples further enhances the smoothing effect, leading to stronger generalization and more robust learning. Across multiple multimodal benchmark datasets, DML consistently outperformed conventional multimodal training. The method improved both classification accuracy and cross-modal retrieval tasks, including matching images with their corresponding text descriptions and retrieving images from textual queries. “Our work provides a theoretical foundation for understanding why multimodal learning consistently outperforms unimodal learning,” the research team said. “It also demonstrates how that understanding can be translated into a simple yet effective training strategy that improves both robustness and generalization.” The study's first author is Jaejun Lee, a researcher in the UNIST Graduate School of Artificial Intelligence. The work has been accepted for presentation at the International Conference on Machine Learning (ICML) 2026, one of the world's leading conferences on artificial intelligence, to be held in Seoul from July 6–11. The research was supported by the National Research Foundation of Korea (NRF) and the Institute for Information & Communications Technology Planning &Evaluation (IITP) through programs funded by the Ministry of Science and ICT. Journal Reference Jae-Jun Lee and Sung Whan Yoon, "Understanding Multimodal Learning: A Loss Landscape Smoothness Perspective," ICML '26 ., (2026).

Abstract The 2025 Los Angeles wildfires highlighted growing risks from climate-driven extremes and the need for reliable wildfire forecasting beyond short lead times. Although the European Center for Medium-Range Weather Forecasts provides global fire weather index forecasts, these products are primarily designed for long-range climate outlooks, and their practical effectiveness remains uncertain in regions with limited local forecasting infrastructure. Here we show that a global deep learning framework, forecasting daily fire weather index values up to 31 days ahead, consistently improves forecast accuracy and reduces bias relative to operational numerical forecasts. The framework learns nonlinear and lagged fire–weather relationships by integrating past fire weather index dynamics and future meteorological conditions. Importantly, forecast bias is reduced in 85% of grid cells where high wildfire exposure coincides with high socioeconomic vulnerability. This demonstrates that data-driven forecasting can help bridge critical information gaps in underserved regions and support more equitable climate risk management. Reliable wildfire forecasts become increasingly difficult beyond the first few weeks, limiting their value for medium-range decision-making. A research team, led by Professor Jungho of the Department of Civil, Urban, Earth, and Environmental Engineering at UNIST, has shown that a global deep learning framework can extend reliable daily wildfire forecasts to 31 days while reducing forecast bias compared with existing operational forecasting systems. Known as FWI-Net, the framework predicts daily values of the Fire Weather Index (FWI), a widely used measure of wildfire danger. Unlike conventional forecasting methods, it combines historical fire-weather conditions with future meteorological forecasts, allowing it to capture the cumulative effects of prolonged heat and drought on future wildfire risk. Across the full 31-day forecasting period, FWI-Net reduced prediction error by 6.6% compared with the operational forecasting system of the European Center for Medium-Range Weather Forecasts (ECMWF). During the first week, prediction error decreased by 12.4%. The model also extended the period of meaningful forecasts under very high wildfire danger by five days. The model performed particularly well in regions where high wildfire exposure coincides with high socioeconomic vulnerability. Forecast bias was reduced across 85% of these areas, and in regions with limited forecasting infrastructure, FWI-Net maintained useful forecasting skill for an average of 22 days. “By combining historical fire-weather conditions with future weather forecasts, the model captures patterns that conventional forecasting systems often miss,” the research team said. “Most importantly, it provides more reliable forecasts for regions facing high wildfire risk despite limited forecasting infrastructure.” “As climate change increases wildfire risk around the world, reliable forecasting is becoming essential for disaster preparedness,” said Professor Im. “We expect this framework to support medium-range wildfire planning while helping reduce information gaps in regions with limited forecasting capacity.” The study was co-led by Professor Yoojin Kang of Kookmin University and Sihyun Lee of UNIST, who served as first authors. The findings were published online in Communications Earth & Environment on June 28, 2026. The research was supported by the Ministry of Environment (ME), the Korea Forest Service, and the National Research Foundation of Korea (NRF). Journal Reference Yoojin Kang, Sihyun Lee, Dongjin Cho, and Jungho Im, "Deep learning-based forecasting provides a pathway to closing wildfire information gaps in underserved regions," Commun. Earth Environment. , (2026).

Abstract We demonstrate that in knowledge distillation for diffusion models, the teacher network's highly complex denoising process—stemming from its substantially larger capacity—poses a significant challenge for the student model to faithfully mimic. To address this problem, we propose a coarse-to-fine distillation framework with LInear Fitting-based distillation (LIFT) and Piecewise Local Adaptive Coefficient Estimation (PLACE). First, LIFT decomposes the objective into a coarse'' alignment and afine'' refinement. The student is then trained on coarse alignment before proceeding to hard refinement. Second, LInear Fitting-based distillation extends LIFT to address spatially non-uniform errors by partitioning outputs into error-based groups, providing locally adaptive guidance. Our comprehensive experimental results demonstrate that ours, \core~with \pick, outperforms previous knowledge distillation on diffusion models based on both U-Net and DiT architectures. Furthermore, as compression rates become exceedingly high, conventional knowledge distillation fails to provide sufficient guidance, thereby preventing lightweight diffusion models from achieving stable training. In contrast, our method demonstrates stable convergence even under such extreme compression ratios. Compact generative AI models promise to bring image generation to personal devices, but reducing model size often comes at the expense of image quality and training stability. Researchers at UNIST have developed a new knowledge distillation framework that enables lightweight diffusion models to learn more effectively from much larger models without increasing inference costs. Led by Professor Jaejun Yoo of the Graduate School of Artificial Intelligence, the study addresses a key challenge in knowledge distillation—the process of transferring the capabilities of a large AI model to a smaller one. The researchers found that as teacher models become increasingly capable, the complexity of their denoising process makes it progressively more difficult for compact student models to reproduce faithfully. Rather than asking the student model to reproduce every detail from the outset, the framework restructures learning into successive stages, allowing it to first capture an image's overall structure before progressively refining finer details. The approach further improves learning by providing additional guidance where prediction errors are greatest while adapting supervision throughout training. This strategy is implemented through two complementary techniques, LIFT (LInear Fitting-based Distillation) and PLACE (Piecewise Local Adaptive Coefficient Estimation). The approach changes only the training strategy, requiring neither additional model parameters nor extra computation during image generation. Across multiple image-generation benchmarks, the proposed framework consistently outperformed existing knowledge distillation methods. It proved effective for text-to-image synthesis, class-conditional generation, and unconditional image generation, while remaining compatible with multiple diffusion architectures, including U-Net, DiT, and MMDiT, the architecture used in Stable Diffusion 3. The advantages became particularly evident under extreme model compression, where the widening gap between teacher and student models often leaves conventional knowledge distillation unable to provide sufficient guidance for stable learning. When compressing a teacher model containing 78.7 million parameters into a student model with just 1.3 million parameters, the proposed method maintained stable convergence and generated substantially higher-quality images. “Rather than modifying model architectures, our approach improves how compact models learn from larger ones,” said Professor Yoo. “Because it introduces no additional model parameters or inference-time computation, it can be readily applied across a wide range of generative AI models, making compact generative AI practical across a wide range of computing environments.” Hyunsoo Han and Sangyeob Yeo of the Graduate School of Artificial Intelligence at UNIST contributed to the study. The findings have been accepted for presentation at the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2026. Journal Reference Hyunsoo Han, Sangyeop Yeo, and Jaejun Yoo, "LIFT and PLACE: A Simple, Stable, and Effective Knowledge Distillation Framework for Lightweight Diffusion Models," CVPR'26

Abstract Cooperation is central to multi-agent reinforcement learning (MARL), yet learned coordination can be fragile when external perturbations disrupt inter-agent interactions. Prior robust MARL methods have primarily considered value-oriented attacks, leaving a gap in robustness when interaction structures themselves are corrupted. In this paper, we propose an interaction-breaking adversarial learning (IBAL) framework that takes an information-theoretic view to construct attacks that impede coordination by perturbing agents' observations and actions, and trains agents to perform reliably under such disruptions. Empirically, our approach improves robustness over existing robust MARL baselines across diverse attack settings and yields stronger performance even under agent-missing scenarios. From autonomous drone swarms to industrial robots and self-driving vehicles, many AI systems depend on multiple agents working together. Yet when communication fails or individual agents become unavailable, coordination can quickly break down. Researchers at UNIST have developed a new reinforcement learning framework that trains AI agents to adapt to these disruptions, enabling them to continue working together under conditions that would typically cause conventional systems to fail. Led by Professor Seungyul Han of the Graduate School of Artificial Intelligence, the research team developed Interaction-Breaking Adversarial Learning (IBAL) , a multi-agent reinforcement learning framework that deliberately disrupts coordination during training. Rather than exposing agents only to ideal conditions, IBAL repeatedly simulates communication failures and missing teammates, encouraging the remaining agents to reorganize their roles and develop new cooperative strategies. Unlike previous robust reinforcement learning methods, which primarily strengthen individual agents against noisy observations or unfavorable actions, IBAL focuses on the interactions between agents. During training, the framework identifies information that is essential for coordination and selectively disrupts it, preparing AI systems to maintain cooperation even when communication becomes unreliable. "Existing robust multi-agent reinforcement learning methods mainly interfere with individual agents," said Sunwoo Lee, the study's first author. “IBAL instead challenges the interactions that enable cooperation, allowing the remaining agents to discover new ways of working together when communication is interrupted or some agents fail.” The researchers evaluated the proposed IBAL on the StarCraft II Multi-Agent Challenge (SMAC). In scenarios where some allied units suddenly became unavailable, conventional models often experienced cascading failures, with some achieving win rates as low as 13.3%. By contrast, agents trained with IBAL adapted their strategies in real time—reorganizing team formations and redistributing roles—and achieved a win rate of 87.0%. “Many real-world AI systems rely on multiple agents working together, making them vulnerable to equipment failures or communication loss,” said Professor Han. “Our framework trains AI agents to adapt to these unexpected disruptions and continue accomplishing their task, providing a foundation for safer and more reliable multi-agent systems.” The study has been accepted for presentation at the International Conference on Machine Learning (ICML) 2026, which will be held in Seoul from July 6 to 11, 2026. The research was supported by the Ministry of Science and ICT (MSIT), the Institute for Information & Communications Technology Planning & Evaluation (IITP), and the National Research Foundation of Korea (NRF). Journal Reference Sunwoo Lee, Mingu Kang, Yonghyeon Jo, and Seungyul Han, "Interaction-Breaking Adversarial Learning Framework for Robust Multi-Agent Reinforcement Learning," ICML '26 , (2026).

Abstract Next-generation 6G communication systems necessitate hardware that simultaneously manages high-frequency signal routing and massive matrix-level computations under extreme constraints. However, conventional digital architectures suffer from high power dissipation and area overhead. Here, we report an oxidized molybdenum disulfide (MoS2)-based memristive platform that integrates nonvolatile radio-frequency (RF) switching and energy-efficient vector–matrix multiplication. By implementing a controlled thermal oxidation process, we achieve stable resistive switching with low switching energy and zero-static power consumption, while maintaining a high cutoff frequency of 33.2 THz, outperforming existing phase-change and microelectromechanical systems (MEMS) technologies. Using system-level simulations, we demonstrate the robustness of this hardware through successful 1024-quadrature amplitude modulation (1024-QAM) demodulation, spectral analysis, and multiple-input multiple-output (MIMO) signal reconstruction, even when accounting for intrinsic device non-idealities. This monolithic integration of high-frequency switching and analog computing provides a scalable solution for energy-efficient deployment of intelligent wireless systems. Future 6G networks and satellite communication systems will require hardware that can process growing amounts of data while operating under strict power and space constraints. Professor Myungsoo Kim and his research team from the Department of Electrical Engineering at UNIST introduced a new semiconductor device that combines high-frequency signal switching and computation on a single platform, offering a more compact and energy-efficient approach to next-generation wireless systems. At the heart of the platform is a multifunctional memristor based on oxidized molybdenum disulfide (MoS2), a two-dimensional semiconductor. The device functions as both a radio-frequency (RF) switch and hardware for in-memory matrix operations, bringing communication and computation onto a single chip. Conventional communication systems typically separate these functions, requiring signals to move between switching circuits and processors. This data movement increases power consumption, processing delays, and chip size. By integrating both functions into a single device, the new platform simplifies signal processing while improving energy efficiency. The memristor also operates without standby power. Once switched between its ON and OFF states, it retains that state even after power is removed, making it particularly well suited to energy-constrained applications such as satellite communications and future 6G infrastructure. The device demonstrated a switching energy of just 140 picojoules per operation and switching power below one milliwatt. It maintained stable operation for more than 40,000 seconds and over 1,000 switching cycles, while achieving a calculated cutoff frequency of 33.2 THz. Experimental measurements verified high-frequency switching up to 67 GHz. To evaluate its computing capability, the researchers performed system-level simulations of wireless communication tasks. The device successfully carried out 1024-quadrature amplitude modulation (1024-QAM) demodulation and multiple-input multiple-output (MIMO) signal reconstruction, demonstrating that a single hardware platform can efficiently support both communication and signal processing. "Commercial RF switches continuously consume standby power and often suffer signal loss at high frequencies," said Juhyo Son, the study's first author. "Our device addresses these limitations while remaining competitive with existing phase-change memory and MEMS technologies in energy efficiency, operating speed, and scalability." Professor Kim added, "Our study demonstrates that oxidized MoS₂ memristors can function not only as high-performance RF switches but also as hardware for the in-memory matrix operations required for next-generation wireless communications. By combining nonvolatile operation, low power consumption, and high-frequency performance, this platform could contribute to more compact and energy-efficient systems for satellite communications, radar, defense electronics, and future 6G RF front ends." The research was supported by the National Research Foundation of Korea (NRF) through the Ministry of Science and ICT (MSIT), the Institute for Information & Communications Technology Planning & Evaluation (IITP), the Space-K BIG Project, and a regional talent development program. The findings were published online in Advanced Functional Materials on June 6, 2026. Journal Reference Juho Son, Changwoo Pyo, Sungmoon Park, et al., “ Oxidized MoS2-Based Multifunctional Memristive Hardware for Energy-Efficient mmWave Signal Processing and In-Memory Matrix Multiplication ,” Adv. Funct. Mater., (2026).