Treffer: Initial-Key Cache: An Efficient KV Cache Strategy Focusing on Initial and Key Tokens for LLMs
Title:
Initial-Key Cache: An Efficient KV Cache Strategy Focusing on Initial and Key Tokens for LLMs
Authors:
Source:
2025 International Joint Conference on Neural Networks (IJCNN) Neural Networks (IJCNN), 2025 International Joint Conference on. :1-8 Jun, 2025
Relation:
2025 International Joint Conference on Neural Networks (IJCNN)
Database:
IEEE Xplore Digital Library