Not All Tokens Are Worth Caching: Learning Semantic-Aware Eviction for LLM Prefix Caches

• URL: http://arxiv.org/abs/2605.18825v1
• Date: Tue, 12 May 2026 18:38:24 GMT
• Title: Not All Tokens Are Worth Caching: Learning Semantic-Aware Eviction for LLM Prefix Caches
• Abstract summary: Prefix caching is a key optimization in Large Language Model (LLM) serving.<n>However, its benefit depends critically on the eviction policy as GPU memory is scarce.<n>We show that different token types within a prompt, including system prompts, user queries, tool outputs, model responses, and chain-of-thought reasoning, exhibit up to 756x variation in reuse rates.<n>We present SAECache, a semantic-adaptive prefix cache eviction policy.
• Score: 15.961259469095571
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Prefix caching is a key optimization in Large Language Model (LLM) serving, reusing attention Key-Value (KV) states across requests with shared prompt prefixes to reduce expensive prefill computation. However, its benefit depends critically on the eviction policy as GPU memory is scarce, and existing policies such as LRU largely treat cached blocks uniformly. This view ignores a fundamental property of LLM prompts: not all tokens are equally worth caching. We show that different token types within a prompt, including system prompts, user queries, tool outputs, model responses, and chain-of-thought reasoning, exhibit up to 756x variation in reuse rates, yet no existing eviction policy exploits this signal. In this paper, we present SAECache (Semantic-Adaptive Eviction for prefix caches), a semantic-adaptive prefix cache eviction policy that addresses this gap through three innovations: (1) a multi-queue architecture that routes KV blocks to task-specific queues with tailored priority metrics, capturing both session reuse in multi-turn requests and structural reuse in templated single-turn requests; (2) a semantic-aware token weighting mechanism that learns the reuse value of different token types online through eviction feedback; and (3) a fully adaptive online learning schema for all parameter updates, including log-normal timing parameters, position decay power, queue weights, and meta-parameters, which eliminates manual tuning and enables automatic adaptation to deployment-specific workload characteristics. Through extensive evaluation across heterogeneous workloads, we demonstrate that SAECache achieves 1.4x-2.7x TTFT improvement over production-style baselines, while fixed-parameter alternatives can degrade by up to 2.7x under workload mismatch -- a failure mode our adaptive approach avoids entirely.

Related papers

• Back from the Future: Key-Value Cache Management by Counter-Causal Surprise [67.1056509495879]
  Key-value (KV) cache management through compression and eviction strategies has emerged as an important research direction in recent years.<n>We propose a simple yet effective KV eviction scheme motivated by the insight that past tokens which can be well-predicted from more recent tokens are redundant.<n>We evaluate our strategy on various open-source LLMs and benchmark datasets showing competitive or improved performance over other state-of-the-art methods.
  arXiv  Detail & Related papers  (2026-07-30T02:42:51Z)
• Make Each Token Count: Towards Improving Long-Context Performance with KV Cache Eviction [65.710271475739]
  We introduce a global retention-based KV eviction method that learns each token's future utility under a unified memory budget.<n>We show that our method substantially reduces KV memory while matching or surpassing full-cache inference.<n>These results suggest that learned, globally calibrated KV eviction is not only a compression technique, but also a mechanism for improving long-context reasoning.
  arXiv  Detail & Related papers  (2026-05-10T16:47:50Z)
• CacheRAG: A Semantic Caching System for Retrieval-Augmented Generation in Knowledge Graph Question Answering [5.427221152449008]
  CacheRAG is a systematic cache-augmented architecture for Knowledge Graph Question Answering.<n>It transforms stateless planners into continual learners.<n>It significantly outperforms state-of-the-art baselines.
  arXiv  Detail & Related papers  (2026-04-28T23:46:47Z)
• Sparse Prefix Caching for Hybrid and Recurrent LLM Serving [0.7284556903703034]
  We formalize sparse prefix caching as checkpoint placement under a distribution over overlap depths.<n>For use cases where requests share a non-trivial prefix, we show that our method consistently improves the frontier traced by standard asymmetry on real-world data.<n>It preserves exact outputs, does not change the recurrent computation itself or require new recurrent update kernels.
  arXiv  Detail & Related papers  (2026-04-17T09:24:58Z)
• Judge Q: Trainable Queries for Optimized Information Retention in KV Cache Eviction [53.83828564664595]
  Large language models (LLMs) utilize key-value ( KV) cache to store historical information during sequence processing.<n>Current methods for KV cache eviction typically utilize the last window from the pre-filling phase as queries to compute the KV importance scores for eviction.<n>We propose Judge Q, a novel training method which incorporates a soft token list.
  arXiv  Detail & Related papers  (2025-09-13T03:34:12Z)
• Semantic Caching for Low-Cost LLM Serving: From Offline Learning to Online Adaptation [54.61034867177997]
  Caching inference responses allows them to be retrieved without another forward pass through the Large Language Models.<n>Traditional exact-match caching overlooks the semantic similarity between queries, leading to unnecessary recomputation.<n>We present a principled, learning-based framework for semantic cache eviction under unknown query and cost distributions.
  arXiv  Detail & Related papers  (2025-08-11T06:53:27Z)
• EPIC: Efficient Position-Independent Caching for Serving Large Language Models [19.510078997414606]
  Caching improves serving performance by reusing Key-Value vectors across requests.<n>Existing context caching requires exact prefixes across requests.<n>We introduce Position-Independent Caching (PIC), which enables modular reuse of KV vectors regardless of prefixes.<n>We also introduce EPIC, a serving system incorporating our new LegoLink algorithm, which mitigates the inappropriate "attention sink" effect at every document beginning.
  arXiv  Detail & Related papers  (2024-10-20T08:42:29Z)
• Efficient Inference of Vision Instruction-Following Models with Elastic Cache [76.44955111634545]
  We introduce Elastic Cache, a novel strategy for efficient deployment of instruction-following large vision-language models. We propose an importance-driven cache merging strategy to prune redundancy caches. For instruction encoding, we utilize the frequency to evaluate the importance of caches. Results on a range of LVLMs demonstrate that Elastic Cache not only boosts efficiency but also notably outperforms existing pruning methods in language generation.
  arXiv  Detail & Related papers  (2024-07-25T15:29:05Z)
• Training-Free Exponential Context Extension via Cascading KV Cache [49.608367376911694]
  We introduce a novel mechanism that leverages cascading sub-cache buffers to selectively retain the most relevant tokens.<n>Our method reduces prefill stage latency by a factor of 6.8 when compared to flash attention on 1M tokens.
  arXiv  Detail & Related papers  (2024-06-24T03:59:17Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.