Wangsong Yin
Scholar

Wangsong Yin

Google Scholar ID: 9hp2_UsAAAAJ
Peking University
Homepage↗Google Scholar↗
Citations & Impact
All-time
Citations
363
 
H-index
4
 
i10-index
3
 
Publications
6
 
Co-authors
0
 
Contact
CVOpen ↗GitHubOpen ↗
Publications
6 items
OmniTide: Co-Designing Algorithms and Systems for Efficient On-Device Omni-LLM Streaming
Cited
0
CacheRepair: Learning to Repair Cross-Chunk Context in RAG for KV Cache Fusion
2026
Cited
0
EdgeCraft: Automated Model Crafting for Edge IoT
2026
Cited
0
Empowering Hybrid Attention Models on NPUs
2026
Cited
0
Dynamic Sparse Attention on Mobile SoCs
2025
Cited
0
ELMS: Elasticized Large Language Models On Mobile Devices
arXiv.org · 2024
Cited
7
Resume
Academic Achievements
  • Paper accepted at MobiCom'25: 'Elastic On-Device LLM Service' (acceptance rate: 10.3%, 36/348)
  • Paper published at SenSys'24: 'PieBridge: Fast and Parameter-Efficient On-Device Training via Proxy Networks' (acceptance rate: 18.5%, 58/313)
  • Paper accepted at SenSys'26 (to appear): 'LLM as a System Service on Mobile Devices' (acceptance rate: 18.9%, 49/258)
  • Published in IEEE Transactions on Mobile Computing (TMC'24): 'Edgellm: Fast On-Device LLM Inference with Speculative Decoding'
  • Published in ACM Computing Surveys (CSUR'25): 'Resource-Efficient Algorithms and Systems of Foundation Models: A Survey'
  • arXiv preprint: 'Dynamic Sparse Attention on Mobile SoCs' — described as 'the best practice to run sparse attention on heterogeneous mobile SoC'