The Key-Value (KV) cache constitutes a dominant source of memory consumption during large language model (LLM) inference, particularly in long-context and co…
机构:南大
来源:arXiv 2610.18642 | AI4Papers 论文推荐平台