# Global Importance Estimation for KV Cache Eviction

> Source: <https://arxiv.org/abs/2611.24969>
> Published: 2026-11-20 05:00:00+00:00

**Recommendation:** When KV cache memory caps long-context inference, evict tokens by multi-hop Personalized PageRank importance over the attention graph rather than recency or instantaneous attention; the authors held competitive accuracy at up to 90% compression.
More in the playbook "Inference and self-hosting": <https://wpnews.pro/playbooks/inference-serving>

arXiv:2611.24969v1 Announce Type: new 
Abstract: Memory consumption of the key-value (KV) cache remains a central bottleneck for long-context LLM inference. We introduce PRKV, a training-free KV eviction method that estimates multi-hop global token importance by applying Personalized PageRank over attention-induced transition graphs. This enables importance propagation beyond recency or instantaneous attention, without modifying the model. We benchmark PRKV on long-context understanding, sparse retrieval, and chain-of-thought reasoning tasks using multiple LLM families. Under moderate to aggressive compression ratios, including up to 90%, PRKV achieves competitive task performance against established eviction baselines while reducing peak memory usage with low end-to-end overhead. These results demonstrate that multi-hop global influence estimation yields a favorable accuracy--memory trade-off for scalable long-context inference.
