Daewon Choi

daeone0920 [AT] kaist.ac.kr

I am a Ph.D. student at the Algorithmic Intelligence Laboratory (ALIN-LAB), advised by Prof. Jinwoo Shin at KAIST. I received my B.S. in Industrial Management Engineering and Computer Science from Korea University in 2024.

My research focuses on addressing practical problems — such as efficiency and safety — of multimodal and language foundation models. I aim to develop algorithms that make these models reliably deployable under real-world constraints, with a current focus on inference-time techniques that improve efficiency and robustness without costly retraining.

I am also collaborating with Amazon AGI on related directions.

news

May 06, 2026	RLDX-1 Technical Report is out on arXiv!
Jan 25, 2026	HAMLET is accepted to ICLR 2026!
Aug 20, 2025	Two papers accepted to EMNLP 2025 Findings: Think Clearly and Mamba Drafters.
Sep 30, 2024	Adversarial Robustification via Text-to-Image Diffusion Models selected for Oral Presentation (top 2.3%) at ECCV 2024.
Feb 28, 2024	Started integrated M.S. & Ph.D. program at KAIST, advised by Prof. Jinwoo Shin.

selected publications

^*Equal contribution · ^†Corresponding author

Preprint

RLDX-1 Technical Report

Dongyoung Kim^*§, Huiwon Jang^*§, Myungkyu Koo^*, Suhyeok Jang^*, Taeyoung Kim^* , Daewon Choi, ..., Jinwoo Shin^§

arXiv, 2026

arXiv Bib HTML Code

@article{kim2026rldx,
  title = {RLDX-1 Technical Report},
  author = {Kim, Dongyoung and Jang, Huiwon and Koo, Myungkyu and Jang, Suhyeok and Kim, Taeyoung and Kim, Beomjun and Yoon, Byungjun and Jang, Changsung and Choi, Daewon and Han, Dongsu and Lee, Donguk and Kwon, Heeseung and Jeon, Hojin and Kang, Jaehyun and Bae, Jaekyoung and Lee, Jihyuk and Lee, Jimin and Won, John and Ahn, Joonwoo and Park, Junhyeong and Sung, Junyoung and Lee, Kyungmin and Han, Minseong and Yoon, Minsung and Joo, Sejune and Son, Seonil and Park, Seungcheol and Cho, Seunggeun and Moon, Seungjun and Kim, Seungku and Dong, Yonghoon and Cho, Yongjin and Kim, Youngchan and Kim, Chang Hwan and Kim, Dohyeon and Kim, Heecheol and Lee, Heewon and Ahn, Hensen and Ryu, Hyungkyu and Choi, Hyunsoo and Shin, Hyunsoo and Jung, Jaeheon and Kim, Jaewoo and Kim, Jinwook and Chang, Joochul and Kim, Joonsoo and Park, Junghun and Park, Jungwoo and Cho, Junho and Park, Junhyeok and Lee, Junwon and Lee, Kangwook and Kim, Kwanghoon and Choe, Kyoungwhan and Bhadu, Manoj and Oh, Nayoung and Kim, Sangjun and Kim, Sangwoo and Shim, Seunghoon and Kim, Seunghyun and Lee, Seungjun and Ka, Seungyup and Yang, Sungryol and Jung, Wook and Shukla, Yashu and Lee, Yeonjae and Bae, Yeonwoo and Shin, Jinwoo},
  journal = {arXiv preprint arXiv:2605.03269},
  year = {2026},
}

ICLR

HAMLET: Switch Your Vision-Language-Action Model into a History-Aware Policy

Myungkyu Koo^*, Daewon Choi^*, Taeyoung Kim, Kyungmin Lee, Changyeon Kim, Younggyo Seo^†, and Jinwoo Shin^†

In International Conference on Learning Representations, 2026

arXiv Bib HTML

@inproceedings{koo2026hamlet,
  title = {{HAMLET}: Switch Your Vision-Language-Action Model into a History-Aware Policy},
  author = {Koo, Myungkyu and Choi, Daewon and Kim, Taeyoung and Lee, Kyungmin and Kim, Changyeon and Seo, Younggyo and Shin, Jinwoo},
  booktitle = {The Fourteenth International Conference on Learning Representations},
  year = {2026},
  url = {https://openreview.net/forum?id=KcJ9U0x6kO},
}

EMNLP

Think Clearly: Improving Reasoning via Redundant Token Pruning

Daewon Choi, Jimin Lee, Jihoon Tack, Woomin Song, Saket Dingliwal, Sai Muralidhar Jayanthi, Bhavana Ganesh, Jinwoo Shin, Aram Galstyan, and Sravan Babu Bodapati

In Findings of the Association for Computational Linguistics: EMNLP 2025

DOI arXiv Bib Poster

@inproceedings{choi-etal-2025-think,
  title = {Think Clearly: Improving Reasoning via Redundant Token Pruning},
  author = {Choi, Daewon and Lee, Jimin and Tack, Jihoon and Song, Woomin and Dingliwal, Saket and Jayanthi, Sai Muralidhar and Ganesh, Bhavana and Shin, Jinwoo and Galstyan, Aram and Bodapati, Sravan Babu},
  booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2025},
  month = nov,
  year = {2025},
  address = {Suzhou, China},
  publisher = {Association for Computational Linguistics},
  url = {https://aclanthology.org/2025.findings-emnlp.1169/},
  doi = {10.18653/v1/2025.findings-emnlp.1169},
  pages = {21437--21451},
}

EMNLP

Mamba Drafters for Speculative Decoding

Daewon Choi, Seunghyuk Oh, Saket Dingliwal, Jihoon Tack, Kyuyoung Kim, Woomin Song, Seojin Kim, Insu Han, Jinwoo Shin, Aram Galstyan, Shubham Katiyar, and Sravan Babu Bodapati

In Findings of the Association for Computational Linguistics: EMNLP 2025

DOI arXiv Bib Poster

@inproceedings{choi-etal-2025-mamba,
  title = {Mamba Drafters for Speculative Decoding},
  author = {Choi, Daewon and Oh, Seunghyuk and Dingliwal, Saket and Tack, Jihoon and Kim, Kyuyoung and Song, Woomin and Kim, Seojin and Han, Insu and Shin, Jinwoo and Galstyan, Aram and Katiyar, Shubham and Bodapati, Sravan Babu},
  booktitle = {Findings of the Association for Computational Linguistics: EMNLP 2025},
  month = nov,
  year = {2025},
  address = {Suzhou, China},
  publisher = {Association for Computational Linguistics},
  url = {https://aclanthology.org/2025.findings-emnlp.1167/},
  doi = {10.18653/v1/2025.findings-emnlp.1167},
  pages = {21404--21418},
}

ECCV

Adversarial Robustification via Text-to-Image Diffusion Models

Daewon Choi^*, Jongheon Jeong^*, Huiwon Jang, and Jinwoo Shin

In European Conference on Computer Vision (ECCV), 2024

Oral Presentation (top 2.3%)

arXiv Bib Code Poster

@inproceedings{choi2024adversarial,
  title = {Adversarial Robustification via Text-to-Image Diffusion Models},
  author = {Choi, Daewon and Jeong, Jongheon and Jang, Huiwon and Shin, Jinwoo},
  booktitle = {European Conference on Computer Vision (ECCV)},
  year = {2024},
}