Seoklea Cho

cho020218@gmail.com

116, Samseonyo-ro 16-gii

Seongbuk-gu, Seoul

About

I’m Cho Seok-lea, an undergraduate student in the Department of Intelligent Systems and IoT at Hansung University.

My research interests lie in efficient on-device AI, edge inference, AI accelerators, and low-level system optimization

I also research federated learning and trustworthy AI, focusing on reliable, secure in distributed environments.

Research Interests

Energy Efficient Edge Inference
Federated Learning: Distributed Optimization, Attacks and Defenses in Distributed AI and FL
Deep Learning/AI Accelerator

Experiences

Undergraduate Researcher, Prof. Jinwoo Gong / 2025.12 - Present
Research Intern, High-Performance Computing Lab, Chung-Ang University, Korea / 2026.04 - 2027.02
LG Aimers 4th / 2024.01 - 2024.03
BoostCourse Coaching Study 13th / 2023.10 - 2023.11
Hansung University, IT Convergence Engineering (Intelligent System and IoT Track) / 2021.03 - 2027.02

Project

I have built edge AI, mobile, IoT capstone, federated learning, and video search projects using PyTorch, CUDA, ExecuTorch, Kotlin, Swift, and Jetpack Compose.

Award

Hansung University, Internet of Things Track Capstone Design Excellence Award (2nd place)
Hansung University, Intelligence System Track Capstone Design Excellence Award (2nd place)

Recent Publication

ICC
Consistent Low-Rank Dequantization for Gradient Inversion under Low-Bit Quantization

Seoklea Cho and Jinu Gong

2027

Submitted to ICC 2027

Abs Bib PDF

This paper proposes consistent low-rank dequantization as a preprocessing method for gradient inversion attacks under low-bit quantization. It alternates low-rank approximation and projection onto quantization intervals to improve feature recovery from quantized model updates.
@misc{cho2027consistent, title = {Consistent Low-Rank Dequantization for Gradient Inversion under Low-Bit Quantization}, author = {Cho, Seoklea and Gong, Jinu}, year = {2027}, note = {Submitted to ICC 2027}, }
IEIE
Evaluating Generalization and Data Leakage in Large Language Models via Data Perturbation

Taemin Kim and Seoklea Cho

2025

IEIE Conference, 2025

Abs Bib PDF

This study addresses the issue of data leakage in Large Language Model evaluation and proposes a perturbation-based robustness test to assess true generalization ability. Semantically preserved but surface-level altered datasets are constructed through synonym substitution and label/schema renaming, then evaluated across recent small language models and standard benchmarks.
@misc{kim2025llmleakage, title = {Evaluating Generalization and Data Leakage in Large Language Models via Data Perturbation}, author = {Kim, Taemin and Cho, Seoklea}, year = {2025}, note = {IEIE Conference, 2025}, }