Scholar
Hyeonjeong Ha
Google Scholar ID: kpDJ2mgAAAAJ
University of Illinois at Urbana-Champaign
Citations & Impact
Citations
31
H-index
4
i10-index
0
Publications
12
Co-authors
27
list available
Publications
Resume (English only)
Academic Achievements
- - Publications:
- - PARTONOMY: Large Multimodal Models with Part-Level Visual Understanding (NeurIPS 2025 Spotlight)
- - SYNTHIA: Novel Concept Design with Affordance Composition (ACL 2025 Main)
- - MM-PoisonRAG: Disrupting Multimodal RAG with Local and Global Poisoning Attacks (preprint)
- - Visual Backdoor Attacks on MLLM Embodied Decision Making via Contrastive Trigger Learning (preprint)
- - PAPO: Perception-Aware Policy Optimization for Multimodal Reasoning (preprint)
- - Energy-Based Transformers are Scalable Learners and Thinkers (preprint)
- - Constructive Distortion: Multimodal LLMs with Test-Time Image Warping (preprint)
- - Must Read: A Systematic Survey of Computational Persuasion (preprint)
- - Generalizable Lightweight Proxy for Robust NAS against Diverse Perturbations (NeurIPS 2023)
- - Effective Targeted Attacks for Adversarial Self-Supervised Learning (ICML 2023 AdvML-Frontiers’ Workshop)
- - Few-shot Transferable Robust Representation Learning via Bilevel Attacks (NeurIPS 2022 ML Safety Workshop)
- - Awards:
- - 2025.09: One paper accepted to NeurIPS 2025 (Spotlight)
- - 2025.08: Awarded the Capital One Illinois Center for Generative AI Safety, Knowledge Systems, and Cyber Security (ASKS) Fellowship Program
- - 2025.05: One paper accepted to ACL 2025
- - 2023.09: Two papers accepted to NeurIPS 2023 Conference
- - 2023.06: One paper accepted to ICML 2023 AdvML-Frontiers’ Workshop
- - 2022.08: Two papers accepted to NeurIPS 2022 ML Safety Workshop
Research Experience
- - Summer 2025: Machine Learning Research Internship at Apple
Education
- - 2024.08 - Present: Ph.D in Computer Science, UIUC, Advisor: Prof. Heng Ji
- - 2022.02 - 2024.02: M.S. in Graduate School of AI, KAIST, Advisor: Prof. Sung Ju Hwang
Background
- - Research Interests: Visual Perception Capability of Multimodal Large Language Models (MLLMs), Trustworthiness of MLLM-powered Framework (e.g., Embodied Agents, RAG), Model Interpretability & Robustness
- - Currently a PhD student at Blender Lab, University of Illinois, Urbana-Champaign (UIUC), advised by Prof. Heng Ji
