Beomseok Kim
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Beomseok Kim has a total of 10 co-authors. Here are the top co-authors by number of shared publications:
- Dongsuk Jeon
Seoul National University
4 shared publications - Sunwoo Lee
Department of Intelligence and Information, Research Institute for Convergence Science, Inter-University Semiconductor Research Center, Seoul National University, Seoul, South Korea
3 shared publications - Byeungseok Yoo
Department of Intelligence and Information, Research Institute for Convergence Science and Inter-University Semiconductor Research Center, Seoul National University, Seoul, South Korea
1 shared publication - Changyu Seong
Department of Intelligence and Information, Research Institute for Convergence Science, Inter-University Semiconductor Research Center, Seoul National University, Seoul, South Korea
1 shared publication - Hunjun Lee
Hanyang University, Seoul, Republic of Korea
1 shared publication - Hyejun Kim
Yonsei University, Seoul, Republic of Korea
1 shared publication
Latest Publications
- Open Access
- 15 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 80-84
FLUX: Frequency Scaling with Layer-wise Utilization for Energy-Efficient NPU Execution (WIP)
With the widespread adoption of Deep Neural Networks (DNNs), Neural Processing Units (NPUs) are emerging as energy-efficient alternatives to GPUs through parallel processing and high data reuse. However, since diverse deep learning kernels have different memory and computation resource requirements, a utilization imbalance between memory and computation resources often occurs. To address this challenge, we propose FLUX, a frequency-scalable NPU system that applies Dynamic Frequency Scaling (DFS) to the core's frequency based on layer-wise arithmetic intensity. FLUX splits the clocking into an adjustable core domain and a fixed system domain. The core-domain frequency is first estimated via roofline cycle analysis and then refined at runtime by an Energy-Delay Product (EDP)-driven calibration. Evaluation on a Gemmini NPU using a 28nm process technology shows that FLUX achieves 5.7%, 16.5%, and 27.9% EDP improvements on 8×8, 16×16, and 32×32 systolic arrays for ResNet50, respectively, reaching within 1.7% of the oracle optimal frequency assignment.
- 15 June 2026
- Published by Association for Computing Machinery (ACM)
- p. 80-84
Open AccessFLUX: Frequency Scaling with Layer-wise Utilization for Energy-Efficient NPU Execution (WIP)
With the widespread adoption of Deep Neural Networks (DNNs), Neural Processing Units (NPUs) are emerging as energy-efficient alternatives to GPUs through parallel processing and high data reuse. However, since diverse deep learning kernels have different memory and computation resource requirements, a utilization imbalance between memory and computation resources often occurs. To address this challenge, we propose FLUX, a frequency-scalable NPU system that applies Dynamic Frequency Scaling (DFS) to the core's frequency based on layer-wise arithmetic intensity. FLUX splits the clocking into an adjustable core domain and a fixed system domain. The core-domain frequency is first estimated via roofline cycle analysis and then refined at runtime by an Energy-Delay Product (EDP)-driven calibration. Evaluation on a Gemmini NPU using a 28nm process technology shows that FLUX achieves 5.7%, 16.5%, and 27.9% EDP improvements on 8×8, 16×16, and 32×32 systolic arrays for ResNet50, respectively, reaching within 1.7% of the oracle optimal frequency assignment.
- Open Access
- 7 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. PP (15498328) , 1-14
An Area- and Energy-Efficient Point-Based 3-D Object Detection Processor With Feature-Decoupled PointNet for Autonomous Driving
Recent point-based neural networks have shown promising performance in 3D object detection by directly processing raw point clouds without converting them into voxels. However, they exhibit stage-specific computational bottlenecks due to the varying computational demands of sampling, grouping, and neural network operations. This paper presents an area- and energy-efficient processor that accelerates point-based neural networks with low latency through algorithm-hardware co-optimization. The proposed feature-decoupled PointNet enables reusing point features by separating MLPs, reducing MAC operations by up to 67.6% without accuracy loss. In addition, density-aware adaptive block pruning eliminates redundant computations in overly dense regions, significantly lowering the computational costs of sampling and grouping. The proposed 3D object detection processor efficiently parallelizes the grouping process by combining a dual-core architecture with a lightweight prioritized shift register, effectively mitigating the workload imbalance. It also adopts a dual-center fused-layer dataflow that maximizes PE utilization while significantly reducing external memory access. Furthermore, the masked max pooling unit eliminates the need for a separate gathering unit by integrating the gathering step into the pooling logic for higher area and power efficiency. Experimental results show that the design provides real-time 3D object detection with a high throughput of 77.9 frames/s, also exhibiting a high accuracy of 83.11 mAP on the KITTI dataset, surpassing prior arts in both accuracy and throughput. In addition, its low energy consumption of 2.5 mJ/frame and high area efficiency of 24.0 frames/s/mm2 make it particularly well-suited for embedded systems for autonomous driving.
- 7 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. PP (15498328) , 1-14
Open AccessAn Area- and Energy-Efficient Point-Based 3-D Object Detection Processor With Feature-Decoupled PointNet for Autonomous Driving
Recent point-based neural networks have shown promising performance in 3D object detection by directly processing raw point clouds without converting them into voxels. However, they exhibit stage-specific computational bottlenecks due to the varying computational demands of sampling, grouping, and neural network operations. This paper presents an area- and energy-efficient processor that accelerates point-based neural networks with low latency through algorithm-hardware co-optimization. The proposed feature-decoupled PointNet enables reusing point features by separating MLPs, reducing MAC operations by up to 67.6% without accuracy loss. In addition, density-aware adaptive block pruning eliminates redundant computations in overly dense regions, significantly lowering the computational costs of sampling and grouping. The proposed 3D object detection processor efficiently parallelizes the grouping process by combining a dual-core architecture with a lightweight prioritized shift register, effectively mitigating the workload imbalance. It also adopts a dual-center fused-layer dataflow that maximizes PE utilization while significantly reducing external memory access. Furthermore, the masked max pooling unit eliminates the need for a separate gathering unit by integrating the gathering step into the pooling logic for higher area and power efficiency. Experimental results show that the design provides real-time 3D object detection with a high throughput of 77.9 frames/s, also exhibiting a high accuracy of 83.11 mAP on the KITTI dataset, surpassing prior arts in both accuracy and throughput. In addition, its low energy consumption of 2.5 mJ/frame and high area efficiency of 24.0 frames/s/mm2 make it particularly well-suited for embedded systems for autonomous driving.
- Open Access
- 19 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 72 (11) , 1760-1764
An Energy-Efficient 3D Semantic Segmentation Processor With Offset-Wise Weight Quantization
Voxel-based point cloud neural networks have gained significant attention for 3D semantic segmentation due to their effectiveness in processing point clouds. However, the high computational overhead of processing large-scale point clouds and the inherent irregularity of these point clouds hinder the fast and energy-efficient acceleration of segmentation. This brief presents a 3D semantic segmentation accelerator with an offset-wise weight quantization technique that drastically reduces computational complexity. The proposed design introduces a unified voxel search unit that efficiently processes various types of operations needed to capture spatial relationships among irregularly stored voxels. In addition, a dual-mode computing engine combined with a novel workload allocation technique enables highly parallel processing and maximizes processing core utilization. Fabricated in a 28-nm CMOS technology, the proposed processor achieves 10.24 TOPS/W energy efficiency, outperforming prior art.
- 19 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems II: Express Briefs
- Vol. 72 (11) , 1760-1764
Open AccessAn Energy-Efficient 3D Semantic Segmentation Processor With Offset-Wise Weight Quantization
Voxel-based point cloud neural networks have gained significant attention for 3D semantic segmentation due to their effectiveness in processing point clouds. However, the high computational overhead of processing large-scale point clouds and the inherent irregularity of these point clouds hinder the fast and energy-efficient acceleration of segmentation. This brief presents a 3D semantic segmentation accelerator with an offset-wise weight quantization technique that drastically reduces computational complexity. The proposed design introduces a unified voxel search unit that efficiently processes various types of operations needed to capture spatial relationships among irregularly stored voxels. In addition, a dual-mode computing engine combined with a novel workload allocation technique enables highly parallel processing and maximizes processing core utilization. Fabricated in a 28-nm CMOS technology, the proposed processor achieves 10.24 TOPS/W energy efficiency, outperforming prior art.
- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 72 (11) , 6553-6566
CLAT: A Clustering-Based Attention Transformer Accelerator for Low-Latency Text Generation in LLMs
Transformer-based large language models (LLMs) excel in text generation but face challenges like memory bandwidth bottlenecks and large key-value (KV) cache sizes as context lengths grow, impacting low-latency performance. Existing accelerators adopt parallelism, model compression, and sparsity exploitation but often fail to fully utilize token-specific sparsity, limiting their effectiveness for long context lengths. CLAT addresses these issues with a low-overhead clustering algorithm that identifies relevant key vector clusters for each token’s query, omitting less relevant vectors with minimal impact. It optimizes memory bandwidth using routing for single-batch inference and introduces scheduling techniques to reduce attention layer latency. Additionally, CLAT compresses model parameters to 4-bit precision and KV caches to 8-bit precision, supported by a multi-precision MAC structure that avoids extra overhead. Validated on Llama2-7B, OPT-6.7B, and Llama3-8B models, CLAT reduces attention layer latency by up to 88.6% and overall text generation latency by up to 34.9%. It improves single-batch text generation throughput by$1.66\times $ to$2.42\times $ over an A100 GPU, demonstrating significant performance gains.- 10 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Circuits and Systems I: Regular Papers
- Vol. 72 (11) , 6553-6566
CLAT: A Clustering-Based Attention Transformer Accelerator for Low-Latency Text Generation in LLMs
Transformer-based large language models (LLMs) excel in text generation but face challenges like memory bandwidth bottlenecks and large key-value (KV) cache sizes as context lengths grow, impacting low-latency performance. Existing accelerators adopt parallelism, model compression, and sparsity exploitation but often fail to fully utilize token-specific sparsity, limiting their effectiveness for long context lengths. CLAT addresses these issues with a low-overhead clustering algorithm that identifies relevant key vector clusters for each token’s query, omitting less relevant vectors with minimal impact. It optimizes memory bandwidth using routing for single-batch inference and introduces scheduling techniques to reduce attention layer latency. Additionally, CLAT compresses model parameters to 4-bit precision and KV caches to 8-bit precision, supported by a multi-precision MAC structure that avoids extra overhead. Validated on Llama2-7B, OPT-6.7B, and Llama3-8B models, CLAT reduces attention layer latency by up to 88.6% and overall text generation latency by up to 34.9%. It improves single-batch text generation throughput by$1.66\times $ to$2.42\times $ over an A100 GPU, demonstrating significant performance gains.