Hyeonho OH

About Me

Hi there! I'm a Master's student in Computer Science at the University of Southern California, graduating in May 2027. Before USC, I earned my B.Eng. in AI & Software from Gachon University, Korea, in 2025. These days I'm fascinated by robot learning, especially world models, simulation for evaluating robot policies, and vision-language-action (VLA) models.

I'm currently applying for PhD positions starting in Fall 2027, and I'm always happy to connect.

Research Experience

Graduate Research Assistant

University of Southern California

Lira Lab — Advisor: Prof. Erdem Bıyık (Sep. 2025 – Present)

Researching a video world model pipeline to augment play data from human-hand demonstrations.
Researching object-centric representations with auxiliary losses and contrastive learning for spatial invariance.

GLAMOR Lab — Advisor: Prof. Jesse Thomason (Sep. 2025 – Present)

Constructed a simulation benchmark environment to evaluate VLA robustness under diverse perturbation sets.
Researched an asynchronous real-time simulation benchmark for evaluating robot policies under inference latency.

Research Intern

Electronics and Telecommunications Research Institute (ETRI) — Advisor: Prof. Hyung-Il Kim (Summer & Winter 2023)

Led a project on robust banner detection and text recognition affected by distortion, occlusion, and noise.
Engineered a novel multi-language context-aware Scene Text Recognition model leveraging CLIP, prompt tuning, and contrastive learning, achieving higher accuracy than the state-of-the-art model.

Undergraduate Research Assistant

Visual AI Lab, Gachon University — Advisor: Prof. Jungchan Cho (Oct. 2021 – May 2023)

Contributed to a vision AI lab that analyzed human detailed behaviors in videos for situational understanding.
Secured 2nd place in a Korean OCR competition organized by the Ministry of Science and ICT, developing an advanced OCR model.

Publications

(* indicates equal contribution)

P. Song*, H. Oh*, I. Singh, J. Thomason, A. Mandlekar, S. Ancha. "Simulation for Evaluating Robot Policies Should Be Asynchronous and Real-Time." Under review, 2026.
H. Lee*, V.-D. Yun*, H. Oh*, D. Panagou, S. P. Karimireddy. "Robust Multi-Agent LLMs under Byzantine Faults." Workshop at AIWILD, ICML, 2026.
J. Morgan, P. Vijay, H. Oh, J. Song, A. Arora, A. Du, G. Sukhatme, J. Thomason, I. Singh. "Colosseum V2: Benchmarking Generalization for Vision Language Action Models." Under review at RA-L, 2026.
H. Oh, J. S. Yun, Y. S. Bae, H. I. Kim. "Investigating Performance of Text Classification Learned with Multi-language for Scene Text Recognition." Institute of Electronics and Information Engineers (IEIE) Summer Conference, pp. 1058-1061, 2023.