Daechul Ahn

I'm a Ph.D. student at SNU Machine Perception and Reasoning Lab. at Seoul National University advised by Prof. Jonghyun Choi.

My research focuses on building embodied AI systems that perceive, reason, and act like humans, progressing from video-language understanding and multimodal alignment to multi-agent reasoning and real-world robotic manipulation.

📬 Open to research internship opportunities — feel free to reach out or connect!

CV / Email / Google Scholar / Semantic Scholar / DBLP / LinkedIn / Github

News

2026.01: BINDER is accepted to ICRA 2026! 🇦🇹
2026.01: LWE is accepted to EACL 2026! 🇲🇦
2025.11: VECTOR is accepted to WACV 2026! 🇺🇸
2025.06: Selected as Outstanding Reviewer at CVPR 2025 (Top 5.6%)
2025.07: HIMA is accepted to COLM 2025! 🇨🇦
2025.02: MA-VR is published in IEEE ACCESS 2025
2024.12: ISR-DPO is accepted to AAAI 2025! 🇺🇸
2024.05: VLM-RLAIF is accepted to ACL 2024 as Oral! 🇹🇭
2023.07: CMOTA is accepted to ICCV 2023! 🇫🇷
2022.10: Received Best Paper Award at the 1st Yonsei AI Workshop
2021.07: PSVL is accepted to ICCV 2021 as Oral! 🌐

Publications

[C] Conference, [J] Journal, [P] Preprint | * denotes equal contribution

Preprints / Under Review

[P1] A Paper about Uncertainty-driven Inference for VLA Models
Daechul Ahn , ...
Under Review, 2026

Current Research: Artificial Intelligence

	[C8] BINDER: Instantly Adaptive Mobile Manipulation with Open-Vocabulary Commands Seongwon Cho* , Daechul Ahn* , Donghyun Shin , Hyeonbeom Choi , San Kim , Jonghyun Choi ICRA 2026 (To appear) [paper (arXiv)] [bibtex] [project] tl;dr: We tackle open-vocabulary mobile manipulation in changing scenes, by proposing BINDER: separate deliberative planning from continuous video monitoring to update state and trigger replanning.
	[C7] Becoming Experienced Judges: Selective Test-Time Learning for Evaluators Seungyeon Jwa , Daechul Ahn , Reokyoung Kim , Dongyeop Kang , Jonghyun Choi EACL 2026 (To appear) [paper (arXiv)] [bibtex] [code] tl;dr: We tackle brittle LLM-as-a-judge evaluation, by proposing Learning While Evaluating: an evolving meta-prompt that self-improves at test time, updated selectively on inconsistent cases.
	[C6] What Happens When: Learning Temporal Orders of Events in Videos Daechul Ahn* , Yura Choi* , Hyeonbeom Choi* , Seongwon Cho , San Kim , Jonghyun Choi WACV 2026 (To appear) [paper (arXiv)] [bibtex] [project] tl;dr: We tackle weak temporal-order understanding in video models, by proposing VECTOR (a temporal-order diagnostic) and MECOT (event-by-event instruction + CoT) to teach explicit ordering.
	[C5] Society of Mind Meets Real-Time Strategy: A Hierarchical Multi-Agent Framework for Strategic Reasoning Daechul Ahn* , San Kim* , Jonghyun Choi COLM 2025 [paper] [bibtex] [code] [project] tl;dr: We tackle long-horizon strategy in real-time games, by proposing a hierarchical multi-agent system where specialist agents propose plans and a meta-planner composes them.
	[J1] Moment-Aware Video Retrieval for Video Corpus Moment Retrieval Yura Choi* , Daechul Ahn* , Jonghyun Choi IEEE ACCESS 2025 [paper] [bibtex] tl;dr: We tackle suboptimal decomposed VCMR learning, by proposing moment-aware video retrieval that ties video retrieval to predicted moments and curriculum hard-negative mining.
	[C4] ISR-DPO: Aligning Large Multimodal Models for Videos by Iterative Self-Retrospective DPO Daechul Ahn* , Yura Choi* , San Kim , Youngjae Yu , Dongyeop Kang , Jonghyun Choi AAAI 2025 [paper] [arXiv] [bibtex] [code] [project] tl;dr: We address preference optimization drift in video LMMs, by proposing self-retrospective DPO that repeatedly re-checks visual evidence to reduce language-only shortcuts.
	[C3] Tuning Large Multimodal Models for Videos using Reinforcement Learning from AI Feedback Daechul Ahn , Yura Choi , Youngjae Yu , Dongyeop Kang , Jonghyun Choi ACL 2024 (Oral presentation) [paper] [bibtex] [code] [project] [demo] tl;dr: We tackle video–language model alignment with limited human feedback, by proposing RLAIF with a context-aware reward that makes AI-generated preferences more video-grounded.
	[C2] Story Visualization by Online Text Augmentation with Context Memory Daechul Ahn , Daneul Kim , Gwangmo Song , Seung Hwan Kim , Honglak Lee , Dongyeop Kang , Jonghyun Choi ICCV 2023 [paper] [bibtex] [code] [project] tl;dr: We address long-range consistency in story visualization, by proposing a Transformer with context memory plus online text augmentation for robust, coherent generation.
	[C1] Zero-shot Natural Language Video Localization Jinwoo Nam* , Daechul Ahn* , Dongyeop Kang , Seong Jong Ha , Jonghyun Choi ICCV 2021 (Oral presentation) [paper] [bibtex] [code] tl;dr: We tackle video moment localization without paired labels, by proposing pseudo-supervision that converts unpaired text and unlabeled videos into synthetic query–segment training pairs.

Past Research: Semiconductor [Show details]

Research conducted at KAIST during M.S. (2015–2017).

First-Author Journal Papers: 3 | Co-Author Journal Papers: 9 | Conference Papers: 5

First-Author Journal Papers

D. Ahn, B.-H. Lee, M.-H. Kang, J. Hur, T. Bang, Y.-K. Choi, "Impact of Crystalline Damage on a Vertically Integrated Junctionless Nanowire Transistor," Appl. Phys. Lett., 2016.
B.-H. Lee*, D. Ahn*, M.-H. Kang, S.-B. Jeon, Y.-K. Choi, "Vertically Integrated Nanowire-based Unified Memory," Nano Lett., 2016.
D. Ahn, M.-L. Seol, J. Hur, D.-I. Moon, B.-H. Lee, J.-W. Han, J.-Y. Park, S.-B. Jeon, Y.-K. Choi, "Ultra-Fast Erase Method of SONOS Flash Memory by Instantaneous Thermal Excitation," IEEE Electron Device Lett., 2015.

Co-Author Journal Papers

M.-S. Kim, D. Ahn, et al., "Electro-Thermal Erasing at 10⁴-Fold Faster Speeds in Charge-Trap Flash Memory," IEEE Electron Device Lett., 2018.
H. Bae, ..., D. Ahn, et al., "Improved Technique for Extraction of Effective Mobility...," J. Nanosci. Nanotechnol., 2017.
T. Bang, B.-H. Lee, ..., D. Ahn, et al., "Low-Frequency Noise Characteristics in SONOS Flash Memory...," IEEE Electron Device Lett., 2017.
B.-H. Lee, M.-H. Kang, D. Ahn, Y.-K. Choi, "Vertically Integrated Nanowire-Based Zero-Capacitor DRAM," ECS J. Solid State Sci. Technol., 2016.
B.-H. Lee, J. Hur, M.-H. Kang, T. Bang, D. Ahn, et al., "A Vertically Integrated Junctionless Nanowire Transistor," Nano Lett., 2016.
J. Hur, B.-H. Lee, M.-H. Kang, D. Ahn, et al., "Comprehensive Analysis of Gate-Induced Drain Leakage...," IEEE Electron Device Lett., 2016.
D. Lee, ..., D. Ahn, et al., "Three-Dimensional Fin-Structured Semiconducting Carbon Nanotube Network Transistor," ACS Nano, 2016.
H. Bae, ..., D. Ahn, et al., "Physically Transient Memory on a Rapidly Dissoluble Paper...," Sci. Rep., 2016.
B.-H. Lee, M.-H. Kang, D. Ahn, et al., "Vertically Integrated Multiple Nanowire Field Effect Transistor," Nano Lett., 2015.

Conference Papers

D. Kim, D. Ahn, M. G. Allen, Y.-K. Choi, "Triboelectrification Driven Fin-FACT...," IEEE MEMS, 2017.
D. Lee, ..., D. Ahn, et al., "First Demonstration of a Wrap-Gated CNT-FET...," IEEE IEDM, 2016.
B.-H. Lee, D. Ahn, M.-H. Kang, S.-B. Jeon, Y.-K. Choi, "Vertically Integrated ZRAM...," 230th ECS Meeting, 2016.
B.-H. Lee, D. Ahn, et al., "Vertically Integrated Multiple Nanowire FET," 228th ECS Meeting, 2015.
B.-H. Lee, M.-H. Kang, J. Hur, D. Ahn, et al., "An Optimum Strategy for the Low Voltage Operation of the Mechanical Switch," IEEE NANO, 2015. (Best Poster Paper Award)

Awards

outstanding reviewer

5.6%

Computer vision Best Paper Award at the 1st Yonsei AI Workshop @ Oct 2022

Academic Services

Reviewer (conference and journal)

CVPR, ICCV, ECCV, AAAI, WACV, ICRA, IJCV, TPAMI, Pattern Recognition

Misc.

I enjoy photography as a hobby. Check out my portfolio at dafoto.info.

This template is from Jon Barron.