From 300KB to 69KB per Token: How LLM Architectures Solve the KV Cache Problem(news.future-shock.ai)138 points by future-shock-ai 7 months ago | 11 comments