Jinhee Kim

first.last at duke.edu

Hi! I’m a PhD student in the CEI Group at Duke ECE, where I’m fortunate to be advised by Prof. Yiran Chen. Before coming to Duke, I was part of the IRIS Lab at Sungkyunkwan University.

My research interests include Efficient Machine Learning, Adaptive Deep Learning Inference, and Model Quantization.

selected publications

MoBiQuant: Mixture-of-Bits Quantization for Token-Adaptive Any-Precision LLM

Dongwei Wang^*, Jinhee Kim^*, Seokho Han^*, and 8 more authors

arXiv preprint arXiv:2602.20191, 2026

Code
Efficient Multi-bit Quantization Network Training via Weight Bias Correction and Bit-wise Coreset Sampling

Jinhee Kim^*, Jae Jun An^*, Kang Eun Jeon, and 1 more author

Advances in Neural Information Processing Systems, 2026

Code
TruncQuant: Truncation-Ready Quantization for DNNs with Flexible Weight Bit Precision

Jinhee Kim^*, Seoyeon Yoon^*, Taeho Lee, and 3 more authors

In 2025 IEEE/ACM International Symposium on Low Power Electronics and Design (ISLPED), 2025

Code

news

Sep 24, 2026	MoBiQuant is accepted at *NeurIPS 2026*. MoBiQuant is a novel any-precision Mixture-of-Bits quantization framework that adjusts weight precision for flexible LLM inference based on token sensitivity!
Oct 10, 2025	I’m also very grateful to receive the *NeurIPS 2025 Scholar Award*.
Sep 18, 2025	EMQNet is accepted at *NeurIPS 2025*. EMQNet reduces the training time of multi-bit networks by up to ~7.88x!
Aug 30, 2025	Very excited to join Duke ECE as a PhD student in Fall ‘25!
May 19, 2025	TruncQuant is accepted at *ISLPED 2025*. TruncQuant enables low-bit inference via simple truncation (bit-shifting) of LSBs. No quantization function, no floating-point operations, and no memory movement—just lightweight integer logic!