I am currently an MPhil student at HKUST(GZ), where I am advised by Prof. Xuming Hu. Concurrently, I am fortunate to conduct research on AI Governance under the guidance of Prof. Mingxun Zhou at HKUST. Previously, I obtained my B.S. in Data Science from Tongji University, under the supervision of Prof. Zhihua Wei.

My research interests lie at the intersection of Multimodal Learning and Trustworthy AI, with a specific focus on generative models, model watermarking, and interpretability. Beyond research, I am deeply passionate about mathematics, programming, and data visualization. Previously I also have the privilege of collaborating with distinguished researchers, including Yibo Yan, Kaichen Huang, and Na Min An.

📢 I am actively seeking academic collaborations! If you are interested in my research, please feel free to reach out at jiahaohuotj@gmail.com.

🔥 News

2026

• 2026.02: 🚀 Code of CausalEmbed released.
• 2026.02: 🌎 Arriving in Chicago as visting student, advised by Prof. Philip S. Yu.
• 2026.01: 🏆 One paper accepted by ICLR’26!
• 2026.01: 📝 One paper submitted to ICML’26.

2025

• 2025.09: 🚀 Code of PMark released.
• 2025.09: 📝 One paper submitted to ICLR’26.
• 2025.07: 🤝 Glad to contribute to MemOS (GitHub 4.6k Stars)!
• 2025.06: 🚀 Code of MMUnlearner released.
• 2025.05: 🏆 Two papers accepted by ACL’25 Findings; One paper accepted by ACL’25 (Industry) Oral!
• 2025.02: 📝 One paper submitted to ACL’25.

2024

• 2024.09: 🚀 Code of MMNeuron released.
• 2024.09: 🏆 One paper accepted by EMNLP’24!
• 2024.06: 📝 One paper submitted to EMNLP’24.

📝 Publications

Preprint

CausalEmbed: Auto-Regressive Multi-Vector Generation in Latent Space for Visual Document Embedding

Jiahao Huo, Yu Huang, Yibo Yan, Ye Pan, Yi Cao, Mingdong Ou, Philip S. Yu, Xuming Hu

Project

• Auto-regressive embedding generation for multi-vector visual document retrieval.
• 30-155x reduction in token count and test-time sclaing for retrieval performance.

ICLR'26

PMark: Towards Robust and Distortion-free Semantic-level Watermarking with Channel Constraints

Jiahao Huo, Shuliang Liu, Bin Wang, Junyan Zhang, Yibo Yan, Aiwei Liu, Xuming Hu, Mingxun Zhou

Project

• Novel and unified theoretical framework on semantic-level watermarking.
• Distortion-free and robust semantic-level watermarking for LLMs.

ACL'25 Findings

MMUnlearner: Reformulating Multimodal Machine Unlearning in the Era of Multimodal Large Language Models

Jiahao Huo, Yibo Yan, Xu Zheng, Yuanhuiyi Lyu, Xin Zou, Zhihua Wei, Xuming Hu

Project

• Reformulate the task of multimodal MU in the era of MLLMs.
• Aims to erase only the visual patterns associated with a given entity while preserving the corresponding textual knowledge encoded within the original parameters of the language model backbone.
• Develop a novel geometry-constrained gradient ascent method MMUnlearner.

ACL'25 (Industry) Oral

MathAgent: Leveraging a Mixture-of-Math-Agent Framework for Real-World Multimodal Mathematical Error Detection

Yibo Yan, Shen Wang, Jiahao Huo, Philip S. Yu, Xuming Hu, Qingsong Wenpng

Project

• Decomposes error detection into three phases
• Each phase handled by a specialized agent: an image-text consistency validator, a visual semantic interpreter, and an integrative error analyzer.

EMNLP'24 main

MMNeuron: Discovering Neuron-Level Domain-Specific Interpretation in Multimodal Large Language Model

Jiahao Huo, Yibo Yan, Boren Hu, Yutao Yue, Xuming Hu

Project

• Investigating the distribution of domain-specific neurons and the mechanism of how MLLMs process features from diverse domains.

📖 Educations

• 2025.10 - present, MPhil Student, HKUST(GZ).
• 2024.10 - 2025.03, Exchange Student, Technical University of Munich.
• 2021.09 - 2025.07, Undergraduate Student, Tongji University.

💻 Internships

📝 Services

• Conference Reviewer: ICLR 2025, ACL 2025/2026, ICML 2026, SIGIR 2026