Shiwei Liu
Scholar

Shiwei Liu

Google Scholar ID: 73IbXtsAAAAJ
ELLIS Institute Tübingen, MPI for Intelligent Systems, University of Oxford, CS@TU/e
Machine LearningDeep LearningLow-dimentional LearningLLMs
Citations & Impact
All-time
Citations
2,672
 
H-index
27
 
i10-index
48
 
Publications
20
 
Co-authors
11
list available
Resume (English only)
Academic Achievements
  • Multiple papers accepted at top-tier conferences including NeurIPS, ICML, ICLR, ACL, EMNLP, AAAI, Interspeech, and MLSys
  • 3 NeurIPS 2025 papers: 'Curse of Depth in LLMs', 'Layerwise Weight Decay', 'Gated Activation Scaling'
  • 3 ICML 2025 papers: 'Masked Next-Token Prediction', 'How Low-Rank Weights Emerge for LLMs', 'Principal Weights for LLM Fine-tuning'
  • 1 ACL 2025 paper: 'Outlier-weighed Layerwise Sampling Fine-tuning'
  • 3 ICLR 2025 papers: 'Mix Layer Normalization', 'Spike-Aware Adam', 'LLM Composable Interventions'
  • 3 NeurIPS 2024 papers: 'Found in the Middle', 'Sparse 3D Medical', 'Alpha Pruning'
  • 2 EMNLP 2024 papers: 'Layer-skip LLM', 'Is c4 enough for LLM Pruning?'
  • 5 ICML 2024 papers (partial titles listed)
  • 2 Interspeech 2024 papers: 'Sparse Multimodal from Scratch', 'Dynamic Data Pruning for Speech'
  • 1 MLSys 2024 paper: 'Q-Hitter: Quantized-Sparse KV Cache'
  • Organizing academic events: Program Chair for CPAL 2026; co-organizing SLLM and SCOPE workshops at ICLR 2025; organizing SPP workshop at IEEE CAI 2025
  • Area Chair for ICASSP 2025
  • Co-organizing 'Edge LLMs' competition at NeurIPS 2024
Research Experience
  • Principal Investigator (PI) at ELLIS Institute Tübingen
  • Group Leader at Max Planck Institute for Intelligent Systems
  • Former Royal Society Newton International Fellow at University of Oxford
  • Former Junior Research Fellow (JRF) at Somerville College, Oxford
  • Postdoctoral fellow at UT Austin, working with Atlas Wang