At this year’s ICT Paper Competition, innovative research across a broad range of fields—including AI semiconductors, neuromorphic computing, generative AI, mobility, VR/AR, large language model (LLM) systems, and next-generation displays—were selected as award-winning entries. The winning papers were chosen through a comprehensive evaluation based on originality of ideas, theoretical rigor, experimental validation, performance assessment, and potential future impact.
 Jae Seung Woo (Seoul National University)

The Grand Prize was awarded to Jae Seung Woo for his research on ultra-fast, low-power synaptic learning implemented using a 3-transistor (3T) embedded NOR flash memory architecture based on a 28 nm CMOS process. The proposed structure alleviates bottlenecks in conventional memory-based neural networks while enabling stable weight modulation and high energy efficiency at the single-cell level. By demonstrating improvements in both process compatibility and performance, the work received high recognition for its strong applicability to future AI semiconductor and neuromorphic systems.

 Hyoseok Lee (Korea Advanced Institute of Science and Technology)

The Excellence Award recipients Hyoseok Lee and Sohwi Lim identified the fundamental cause of artifacts and quality degradation in existing Latent Diffusion Inverse Solvers, which stems from unstable reverse-diffusion dynamics. To address this issue, their paper proposes a Measurement-Consistent Langevin Corrector (MCLC) that reduces distribution mismatch at each time step. This research presents a new direction that substantially improves both the stability and quality of LDM-based zero-shot inverse solvers.

 Juhee Heo (Sookmyung Women’s University)

Another Excellence Award recipient, Juhee Heo, presented a highly contemporary and impactful study by integrating quantum convolutional neural networks (QCNNs) with reinforcement learning (PPO) to jointly consider detection, transmission, reliability, and freshness. In addition, the proposed predictive-based (proactive) information delivery framework for enhancing Cooperative Adaptive Cruise Control (CACC) safety offers a meaningful research direction with strong potential for further studies in future 6G-based intelligent transportation infrastructures.

 Johnghyun Ko (Seoul National University)

The Distinguished Award recipient Jonghyun Ko was recognized for research that proposes a new direction for generative AI hardware by implementing both probabilistic sampling and deterministic computation within a single device using HfO₂-based ferroelectric tunnel junctions. By realizing latent-space sampling at the hardware level, the study is expected to have significant academic and technological impact on next-generation intelligent semiconductor research, considering its creativity and scalability.

 Minhyuk Kim (Kyung Hee University)

Distinguished Award recipients Minhyuk Kim and Junhyeong Shim addressed the fundamental trade-off in VR streaming between high-resolution delivery and increased encoding latency by leveraging user attention prediction. Rather than relying solely on conventional foveated streaming approaches that consider only gaze location, their work introduces a novel design framework based on attention-probability-driven grid merging and adaptive resolution allocation, greatly enhancing the feasibility of real-world system implementation.

 Jeongwoo Kim (Daegu Gyeongbuk Institute of Science and Technology)

Distinguished Award recipients Jeongwoo Kim and Jaehun Lee reexamined caching strategies for large language model service systems. Existing approaches—such as exact-match, prefix, and semantic caching—are inefficient due to neglecting query semantics or relying on traditional heuristics. Their study proposes a centroid-based caching method to efficiently manage redundant queries, along with a cache replacement strategy that accounts for semantic locality and dynamically adjusts cache hit rates to balance accuracy and latency under varying workloads.

 Hyunwook Cho (Pohang University of Science and Technology)

Distinguished Award recipients Hyunwook Cho, Donghyuk Kim, Yonggon Park, Seunghoon Oh, Sangbu Yoon, and Jaeyong Lee presented an innovative approach to addressing the ECC decoding bottleneck encountered during tiny-read operations in SSDs by reusing Locally Correctable Codes (LCC). Achieving more than a 60% reduction in read latency without requiring additional hardware modifications, the work was highly praised for both its creativity and practicality.

 Hyunsu Jung (Kyungpook National University)

Distinguished Award recipient Hyunsu Jung proposed a geometric-phase-based varifocal lens optical system that enables control over virtual content depth in AR devices. The system supports real-time, full-color AR images with multiple polarizations and depths, effectively addressing the vergence-accommodation conflict. The study includes concrete models and experimental demonstrations for full-color implementation, while minimizing the number of switching elements required for variable depth control.

 Soeun Lee (Yonsei University)

Finally, Distinguished Award recipients Soeun Lee, Seokgyu Hong, Juhyun Lee, and Yongseon Hwang presented a study aimed at resolving the mobility–reliability trade-off of LTPO-grade oxide driving TFTs through a dual-channel structure combining ALD-IGO and sputtered IGZO. The work demonstrates high industrial impact and strong potential for extension to next-generation high-resolution, low-power display backplane devices, with a high level of completeness in both process development and analytical methodology.

Source: https://m.etnews.com/20251216000203
Translated by: Changhoon Kang, English Editor of the Department of Electrical and Computer Engineering, changhoon27@snu.ac.kr