Kazuki Egashira
Scholar

Kazuki Egashira

Google Scholar ID: lCLmXUkAAAAJ
ETH Zurich
Machine LearningSecurity
Citations & Impact
All-time
Citations
65
 
H-index
3
 
i10-index
1
 
Publications
7
 
Co-authors
19
list available
Resume (English only)
Academic Achievements
  • 2025, 'Fewer Weights, More Problems: A Practical Attack on LLM Pruning', published in arXiv
  • 2025, 'Mind the Gap: A Practical Attack on GGUF Quantization', published in ICML and presented orally at BuildingTrust@ICLR25
  • 2024, 'Exploiting LLM Quantization', published in NeurIPS and presented orally at NextGenAISafety@ICML24
  • 2025, IE Special Award (Japanese domestic conference)
  • 2025, The Dean's Award of Graduate School of Information Science and Technology at Univ. of Tokyo
  • 2023, MIRU Interactive Presentation Award (Japanese domestic conference)
Research Experience
  • Conducting research work at SRI Lab, with the position of PhD Student.
Background
  • Since June 2025, pursuing a Ph.D. at the Secure, Reliable, and Intelligent Systems Lab at ETH Zürich, advised by Prof. Martin Vechev.