Deming Chen

A new method speeds up LLM text generation using additional prediction heads.

A new approach enhances federated learning by addressing slow clients effectively.

A new method reduces KV cache size while maintaining high model performance.

This article discusses recent developments to improve efficiency in Large Language Models.