Alvin Jinsung Choi

I am a CS PhD candidate at RSN Lab, which is part of the UT Austin Department of Computer Science and Texas Robotics, advised by Prof. Volkan Isler.

I am interested in Robotics, Computer Vision, and Embodied AI, focusing on building machines that percepts, reason, and interact with the physical world.

Previously, I completed my Bachelor's degree and Master's degree in Electrical Engineering at KAIST. I was fortunate enough to have worked with Prof. Hyun Myung during the years.

Email: alvinjinsung [at] utexas [dot] edu
Office: 2317 Speedway, Austin, Texas, United States of America.

Research Interests

My research goal is to advance robotics system to perceive the physical world in a way that is similar to how humans do. Specifically, I am interested in learning representations of the physical world through uncertainty-aware perception and action, and aligning them with human expectations. To this end, I am currently focusing on the following topics, but not limited to:

Embodied AI
Perception-Action Coupling
Active Perception
Implicit Neural Representations
Neural 3D Reconstruction

News

[Aug 2026] I will be joining the RSN Lab at UT Austin as a CS PhD student, advised by Prof. Volkan Isler!

[Jan 2026] One paper accepted to ICRA 2026! CLUE is a navigation framework that adaptively balances the use of contextual rooms and objects by leveraging commonsense knowledge extracted from an offline LLM for zero-shot object-goal navigation.

[Nov 2025] I defended my master thesis! A huge thanks to my master's committe Prof. Hyun Myung, Prof. Kuk-Jin Yoon, and Prof. Sung-Eui Yoon for their time, feedback, and guidance throughout the thesis process. My thesis is based on my work NeuDonatello which leverages estimated SDF uncertainty to improve 3D surface reconstruction.

[Nov 2025] One paper accepted to ICCAS 2025! GSDB is a lightweight database construction pipeline for Gaussian splatting map-based visual localization. (Selected as Best Paper Award)

[Oct 2025] Participated in IROS 2025!

[Oct 2025] Our Vision-Language-Autonomy model ranked 4th in CMU VLN Challenge, part of IROS 2025 workshop on AI Meets Autonomy!

[Oct 2025] One poster accepted to Humanoids 2025 workshop on Bridging Humanoid Robotics and Foundation Models! ActiveGrounder is a framework that transforms 3D visual grounding from a passive recognition task into an active exploration paradigm.

[Sep 2025] Participated in CoRL 2025 as a student volunteer!

+ More

Publications

NeuDonatello: Uncertainty-Aware SDF Learning for High-Hidelity Neural Surface Reconstruction

Alvin Jinsung Choi, Wanhee Kim, Taeyun Kim, Dasol Hong, Wooju Lee, Hyun Myung
BMVC 2026 (Under Review)
Webpage • Paper (Coming Soon) • Code (Coming Soon)

CLUE: Adaptively Prioritized Contextual Cues by Leveraging a Unified Semantic Map for Effective Zero-Shot Object-Goal Navigation

Taeyun Kim, Alvin Jinsung Choi, Dasol Hong, Hyun Myung
ICRA 2026
Webpage • Paper • Code (Coming Soon)

ActiveGrounder: 3D Visual Grounding with Object-Hull-Guided Active Observation

Dasol Hong, Juhye Park, Taeyun Kim, Jeewon Kim, Jei Kong, Wanhee Kim, Alvin Jinsung Choi, Wooju Lee, Hyun Myung
Humanoids 2025 Workshop on Bridging Humanoid Robotics and Foundation Models
Webpage

GSDB: A Lightweight Database for Gaussian Splatting Map-based Visual Localization Leveraging Edge-aware and Quality-guided View Filtering

Sungjae Shin, Wanhee Kim, Alvin Jinsung Choi, Hyun Myung
ICCAS 2025
Best Paper Award • Webpage • Paper • Code (Coming Soon)

Projects

Vision-Language-Autonomy: CMU VLA Challenge on IROS'25 Workshop on AI Meets Autonomy (4th place)

Dasol Hong, Juhye Park, Taeyun Kim, Alvin Jinsung Choi, Jeewon Kim, Jei Kong, Wooju Lee, Hyun Myung
IROS'25 Workshop on AI Meets Autonomy: CMU VLA Challegne (4th place)
Webpage • Code