Scholar
Haoran Li
Google Scholar ID: O3JRO9AAAAAJ
PhD in HKUST
Citations & Impact
Citations
1,130
H-index
12
i10-index
16
Publications
20
Co-authors
12
list available
Publications
Resume (English only)
Academic Achievements
- Preprints: 'Federated Domain-Specific Knowledge Transfer on Large Language Models Using Synthetic Data' (2024), 'FedCQA: Answering Complex Queries on Multi-Source Knowledge Graphs via Federated Learning' (2024), 'Privacy in Large Language Models: Attacks, Defenses and Future Directions' (2023). Selected Publications: 'PrivaCI-Bench: Evaluating Privacy with Contextual Integrity and Legal Compliance' (to appear at ACL 2025), 'Can Indirect Prompt Injection Attacks Be Detected and Removed?' (to appear at ACL 2025), 'Defense Against Prompt Injection Attack by Leveraging Attack Techniques' (to appear at ACL 2025). Project Manager of the Privacy Checklist Project, aiming to remove the Alignment Tax by building a powerful compliance reasoner for privacy and safety regulations to safeguard foundational models and their applications.
Research Experience
- Intern at Minimax's LLM Alignment team from July to November 2024, working on implementing an instruction hierarchy for their models with enhanced robustness against prompt injection attacks; Intern at Toutiao AI Lab, Bytedance for NLP research during Summer of 2022.
Education
- Ph.D. in Computer Science from the Hong Kong University of Science and Technology, supervised by Prof. Yangqiu Song, awarded on January 31, 2025; B.S. in Computer Science and Math-CS track from the Hong Kong University of Science and Technology, graduated in 2020.
Background
- Research interests: privacy studies in NLP, including privacy attacks and defenses on (Large) Language Models, differential privacy, and contextual integrity. Currently a postdoc at HKUST.
Miscellany
- Recipient of the Jockey Club STEM Early Career Research Fellowship for Translation and Application, grateful for the generous support from the Hong Kong Jockey Club Charities Trust.
