Seungjae Shin

Experience & Education

Qualcomm AI Research · Oct. 2024–Present
Senior ML Research Engineer
Research on model quantization and efficient machine learning.

KAIST
Ph.D., Industrial Engineering · 2020–2024
M.S., Industrial Engineering · 2018–2020
B.S., Industrial Engineering · 2013–2018
Graduate research advisor: Prof. Il-Chul Moon

National University of Singapore · Jan.–May 2017
Visiting Student, Electrical and Computer Engineering

Publications

* Equal contribution.

2026

Pivot Gradient Scaling: Reducing Dwell Time to Stabilize Extreme Low-Bit Quantization-aware Training
Kim, V., Kim, J., Seungjae Shin, Lee, J. H., You, J., Chen, A., and Choi, J.
Findings of EMNLP, 2026 · Paper

2025

Generalized Gumbel-Softmax Gradient Estimator for Generic Discrete Random Variables
Joo, W., Kim, D., Seungjae Shin, and Moon, I.
Pattern Recognition Letters, 196:148–155, 2025 · Paper

Sequential Neural Joint Estimation for Likelihood-Free Inference
Kim, D., Song, K., Seungjae Shin, Kang, W., Moon, I., and Joo, W.
Journal of the Korean Statistical Society, 54(3):858–887, 2025 · Paper

2024

Unknown Domain Inconsistency Minimization for Domain Generalization
Seungjae Shin*, Bae, H.*, Na, B., Kim, Y., and Moon, I.
ICLR, 2024 · Paper · Code

Dirichlet-based Per-sample Weighting by Transition Matrix for Noisy Label Learning
Bae, H., Seungjae Shin, Na, B., and Moon, I.
ICLR, 2024 · Paper

Make Prompts Adaptable: Bayesian Modeling for Vision-Language Prompt Learning with Data-Dependent Prior
Cho, Y., Bae, H., Seungjae Shin, Youn, Y. D., Joo, W., and Moon, I.
AAAI, 2024 · Paper · Code

DPO-Finetuned Large Multi-Modal Planner with Retrieval-Augmented Generation @ EgoPlan Challenge ICML 2024
Lee, K., Kang, M., Na, H., Bae, H., Na, B., Kwon, D., Seungjae Shin, Kim, Y., Kim, T., Yun, S., and Moon, I.
ICML Workshop on Multi-modal Foundation Model Meets Embodied AI, 2024 · Paper

2023

Frequency Domain-based Dataset Distillation
Shin, D.*, Seungjae Shin*, and Moon, I.
NeurIPS, 2023 · Paper · Code

Loss-Curvature Matching for Dataset Selection and Condensation
Seungjae Shin*, Bae, H.*, Shin, D., Joo, W., and Moon, I.
AISTATS, 2023 · Paper · Code

Evaluation of Optimal Scene Time Interval for Out-of-hospital Cardiac Arrest Using a Deep Neural Network
Seungjae Shin*, Bae, H. S.*, Moon, H. J., Kim, G. W., Cho, Y. S., Lee, D. W., Jeong, D. K., Kim, H. J., and Lee, H. J.
American Journal of Emergency Medicine, 63:29–37, 2023 · Paper

2022

Improving Group-based Robustness and Calibration via Ordered Risk and Confidence Regularization
Seungjae Shin, Na, B., Bae, H., Jang, J., Kim, H., Song, K., Cho, Y., and Moon, I.
ICML Workshop on Spurious Correlations, Invariance, and Stability (ICML-SCIS), 2022 · Paper

From Noisy Prediction to True Label: Noisy Prediction Calibration via Generative Model
Bae, H.*, Seungjae Shin*, Na, B., Jang, J., Song, K., and Moon, I.
ICML, 2022 · Paper · Code

Soft Truncation: A Universal Training Technique of Score-based Diffusion Model for High Precision Score Estimation
Kim, D., Seungjae Shin, Song, K., Kang, W., and Moon, I.
ICML, 2022 · Paper · Code

2021

ABC: Auxiliary Balanced Classifier for Class-imbalanced Semi-supervised Learning
Lee, H., Seungjae Shin, and Kim, H.
NeurIPS, 2021 · Paper · Code

Refine Myself by Teaching Myself: Feature Refinement via Self-Knowledge Distillation
Ji, M., Seungjae Shin, Hwang, S., Park, G., and Moon, I.
CVPR, 2021 · Paper · Code

Counterfactual Fairness with Disentangled Causal Effect Variational Autoencoder
Kim, H., Seungjae Shin, Jang, J., Song, K., Joo, W., Kang, W., and Moon, I.
AAAI, 2021 · Paper

2020

Neutralizing Gender Bias in Word Embeddings with Latent Disentanglement and Counterfactual Generation
Seungjae Shin, Song, K., Jang, J., Kim, H., Joo, W., and Moon, I.
Findings of EMNLP, 2020 · Paper

Bivariate Beta-LSTM
Song, K., Jang, J., Seungjae Shin, and Moon, I.
AAAI, 2020 · Paper · Code

FEWER: Federated Weight Recovery
Shin, Y., Lee, G., Seungjae Shin, Yun, S., and Moon, I.
1st Workshop on Distributed Machine Learning (DistributedML), 2020 · Paper

Forecasting the Concentration of Particulate Matter in the Seoul Metropolitan Area Using a Gaussian Process Model
Jang, J., Seungjae Shin, Lee, H., and Moon, I.
Sensors, 20(14):3845, 2020 · Paper

Preprints

Unifying Block-wise PTQ and Distillation-based QAT for Progressive Quantization toward 2-bit Instruction-Tuned LLMs
Lee, J. H.*, Seungjae Shin*, Kim, V., You, J., and Chen, A.
arXiv:2506.09104, 2025 · Preprint

Neural Posterior Regularization for Likelihood-Free Inference
Kim, D., Song, K., Seungjae Shin, Kang, W., Moon, I., and Joo, W.
arXiv:2102.07770, 2021 (revised 2022) · Preprint

Adversarial Likelihood-Free Inference on Black-Box Generator
Kim, D., Joo, W., Seungjae Shin, Song, K., and Moon, I.
arXiv:2004.05803, 2020 · Preprint

Honors & Awards

Qualcomm Innovation Fellowship Korea · Winner, 2023 and 2022
Best Paper Award (Junior), Korean Software Congress · 2017
1st Prize, KAIST Invention Award · 2017
Honor Student Scholarship, KAIST · 2016
Dean’s List, KAIST · 2014

Academic Service & Teaching

Reviewer for NeurIPS, ICML, ICLR, CVPR, EMNLP, and ACL.

Taught computer science to junior students through Data Diving, Daejeon EduSci Center, and Aura Edu.