ShengYun Peng
Scholar

ShengYun Peng

Google Scholar ID: 3557QYYAAAAJ
Georgia Tech
Machine Learning
Citations & Impact
All-time
Citations
769
 
H-index
9
 
i10-index
9
 
Publications
20
 
Co-authors
18
list available
Resume (English only)
Academic Achievements
  • NeurIPS'24: Proposed the first framework (LLM safety basin) explaining alignment collapse from minimal unsafe data during fine-tuning
  • BMVC'23: Developed SOTA adversarially robust CNN architectures on RobustBench; Best Poster Award
  • NeurIPS'25: Introduced a dynamic safety shaping framework for LLM finetuning risk mitigation
  • ACL'25: Conducted scaling study on optimal inference for video vision-language models
  • ICCV'25: Built a scalable framework enabling MLLMs to understand composite images
  • NeurIPS'23 / AAAI'24 / NeurIPS'24: Developed UniTable, a modular table parsing system (470+ GitHub stars, multiple oral presentations)
  • Authored distributed systems tutorials on Medium with 33K+ readers
  • Published papers at top venues including NeurIPS, ICCV, ACL, EMNLP, BMVC, AAAI, and ICLR