Scholar
Wenxuan Zeng
Google Scholar ID: P1c6nDYAAAAJ
Peking University
Citations & Impact
Citations
185
H-index
6
i10-index
5
Publications
13
Co-authors
11
list available
Publications
Resume (English only)
Academic Achievements
- Paper: MPCache: MPC-Friendly KV Cache Eviction for Efficient Private LLM Inference, NeurIPS 2025
- Paper: Towards Efficient Privacy-Preserving Machine Learning: A Systematic Review from Protocol, Model, and System Perspectives, ACM Computing Survey (CSUR) Submission
- Paper: H2EAL: Hybrid-Bonding Architecture with Hybrid Sparse Attention for Efficient Long-Context LLM Inference, ICCAD 2025
- Paper: UniCAIM: A Unified CAM/CIM Architecture with Static-Dynamic KV Cache Pruning for Efficient Long-Context LLM Inference, DAC 2025
- Paper: OptiPrime: Efficient Private Inference at ImageNet Scale, ASPLOS 2025 Submission
- Paper: FlexHE: A Flexible Kernel Generation Framework for Homomorphic Encryption-Based Private Inference, ICCAD 2024
- Paper: PrivQuant: Communication-Efficient Private Inference with Quantized Network/Protocol Co-Optimization, ICCAD 2024
- Paper: BAT: Behavior-Aware Human-Like Trajectory Prediction for Autonomous Driving, AAAI 2024
- Paper: EQO: Exploring Ultra-Efficient Private Inference with Winograd-Based Protocol and Quantization Co-Optimization
Background
- Currently a third-year master student at the Institute for Artificial Intelligence, Peking University (PKU), supervised by Prof. Meng Li and Prof. Runsheng Wang. Research interests primarily focus on efficient AI algorithms and systems, agentic AI, multimodal LLM, and LLM reasoning. Since 2022, he has also explored privacy-preserving machine learning (PPML), focusing on accelerating private inference systems via protocol-algorithm co-optimization.
Miscellany
- Maintains a blog where he records his study notes and knowledge summaries about computer science since 2019 (over 300 blogs and 850,000+ visits).
