Beyond KV Caching: Shared Attention for Efficient LLMs

Arxiv Papers - Beyond KV Caching: Shared Attention for Efficient LLMs