ReCache: Efficient KV Cache Reuse and Compression for Tool-Augmented LLM Agents

工具增强 agent 跨请求重复编码相同工具/技能 schema，不同组合和顺序阻止标准前缀缓存命中

Resource-wise attention：独立缓存每个工具表示的 KV 状态，支持跨请求复用和压缩

显著降低工具增强 agent 的推理计算和内存开销；缓存命中率高于标准前缀缓存

大量工具的实时 agent 推理效率直接影响延迟，与全双工系统的响应时间目标相关

Agentic language models repeatedly encode tool and skill schemas that recur across requests in different combinations and orders, preventing standard prefix caching from reusing their key--value (KV) states. We introduce \textbf{ReCache}, a framework for independently caching resource representations while reducing their inference-time computational and memory overhead. Resource-wise attention removes cross-resource interactions and assigns resource-local positions, producing composition-invaria