发表机构
Stanford University（斯坦福大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究针对KV缓存驱逐规则，通过分析168192个注意力行的最优子集情况，提出无需训练的分配器ContourKV，其在配对比较中表现优于多数现有技术，且与最强基线持平。
AI 中文摘要
每一个已部署的稀疏注意力机制或KV缓存驱逐规则都会保留部分键（key），丢弃其余键，并对保留的键重新归一化注意力权重。对来自五个模型的168192个注意力行进行精确最优子集枚举，结果显示保留最大权重的做法已接近最优，因为最优子集仅缩小了与全注意力之间中位数2%至5%的剩余差距。如果选择带来的改进如此微小，那么已发表的驱逐方法之间的优势必然来自其他方面，因此我们测量每种方法占用的字节数。在共享评估流程中，最强的与查询无关的方法会占用完整缓存，因为它们的每头选择以掩码形式存储，只有不规则的每头存储能释放该内存。对一个固定选择实施名义预算会损失14至62个基准点。我们将87.6点的检索优势归因于在问题可见时计算的排名。ContourKV是一种基于丢弃质量统计量构建的无需训练的分配器，在与现有技术的160次配对比较中赢了93次，在预算强制基线的字节数下输了22次，且与其中最强的方法持平。
英文摘要
Every deployed sparse-attention or KV-cache-eviction rule keeps a subset of the keys, discards the rest, and renormalizes the attention weights over the kept set. Enumerating the exact best subset under that constraint on $168{,}192$ attention rows from five models shows that keeping the largest weights is already near-optimal, since the best subset closes only a median $2$ to $5\%$ of the remaining gap to full attention. If selection closes this little, published margins between eviction methods must come from elsewhere, so we measure the bytes each method holds. In the shared evaluation pipeline, the strongest query-agnostic methods hold the full cache because their per-head selections are stored as masks, and only ragged per-head storage frees that memory. Enforcing a nominal budget on one fixed selection costs $14$ to $62$ benchmark points. We trace an $87.6$-point retrieval margin to rankings computed while the question is visible. ContourKV, a training-free allocator built from the dropped-mass statistic, wins $93$ of $160$ paired comparisons against that state of the art and loses $22$ at the byte count of the budget-enforcing baselines, and it ties the strongest of them.
Comments18 pages; revised wording in 2.3 for increased accuracy (main results unchanged)
