Scholar
Jianwei Li
Google Scholar ID: 6lFRe2MAAAAJ
PhD Student from North Carolina State University
Citations & Impact
Citations
172
H-index
7
i10-index
4
Publications
14
Co-authors
9
list available
Publications
Resume (English only)
Academic Achievements
- Paper 'Safety Alignment Can Be Not Superficial With Explicit Safety Signals' accepted by ICML 2025
- Paper 'Greedy Output Approximation: Towards Efficient Structured Pruning for LLMs Without Retraining' accepted by CPAL 2025
- Paper 'Beyond Gradient and Priors in Privacy Attacks' selected as Oral at FL@FM-NeurIPS 2023
- Two papers accepted by EMNLP 2023
- Paper 'FP8-BERT: Post-Training Quantization for Transformer' published at DCAA@AAAI 2023
- Authored multiple papers on LLM safety, efficient pruning, privacy attacks, and keystroke dynamics authentication
- Served as reviewer for DCAA 2023 and AAAI 2024
- Publicity Chair for KDD 2023 Workshop on Resource-Efficient Learning
- Publicity Chair for AAAI 2023 Workshop on DL-Hardware Co-Design for AI Acceleration
- Passed PhD Written Preliminary Exam at NCSU (Oct 2025)
- Participated in AI Hardware Summit 2022; Moffett S30 Accelerator won MLPerf V2.1
Background
- Third-year PhD student at North Carolina State University, focusing on AI Safety and AI Efficiency
- Specializes in data privacy, adversarial attacks, model robustness and uncertainty, and Large Language Model (LLM) alignment
- Also involved in model compression and red-teaming research
- Actively seeking summer 2026 internship opportunities in Artificial Intelligence, especially in AI Safety and Efficient AI
- Open to industry collaborations in Shadow LLM-related research
