AI Paper Atlas

Research map / Large language models / Reasoning and chain of thought

Intent-Aware Caching for Efficient LLM Serving

Kazi Hasan Ibn Arif, JinYi Yoon, Dimitrios S Nikolopoulos, Hans Vandierendonck, Deepu John, Bo Ji

NeurIPS 2026 · Large language models › Reasoning and chain of thought · approach: KV cache compression · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes an intent-aware caching approach for LLM serving to improve prefix cache hit rates in multi-turn conversations.

Problem

Prefix caching in LLM serving for multi-turn conversations achieves low hit rates because cache reuse depends on whether a conversation continues an earlier prefix, which is hard to predict.

Approach

Uses user intent to guide which conversation prefixes' KV states are cached and reused; details beyond the abstract are not available.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use intent-aware eviction (IAC) instead of LRUSGLang prefix cache · ShareChat 140K tokenscache hit rate: 12.2% -> 34.6%Improved
use IAC instead of learned LPCSGLang prefix cache · WildChat 140K tokenscache hit rate: 17.1% -> 24.8%Improved
IAC instead of LRUQwen-1.5-110B serving · ShareChat 5 req/smean TTFT: 6.98 s -> 3.76 sImproved
IAC instead of LRU at high loadLlama-3-70B serving · ShareChat 10 req/saverage latency: 7.01 s -> 3.41 sImproved
IAC instead of LPC for memory useLlama-3-70B · ShareChatGPU memory for similar hit rate: 51.9 GB -> 24.4 GB (53% saving)Improved
add reuse-probability signal to recency signal (ablation)IAC eviction priority · ShareChat/WildChatcache hit rateImproved
order signals (recency, reuse) instead of (reuse, recency) (ablation)IAC eviction priority · ShareChat/WildChatcache hit rateImproved
add size as final tie-breaker (ablation)IAC eviction priority · ShareChat/WildChatcache hit rateImproved
group by intent plus turn instead of turn only (ablation)IAC statistics · ShareChatcache hit rateImproved
use 30 intent categories instead of coarser 10 or 20 (ablation)IAC · ShareChatcache hit rateImproved
randomize intent labels at serving time (0% to 100%) (ablation)IAC · ShareChat 170K tokenscache hit rate: graceful drop to 30% at full randomizationHurt
use median delay as reuse window instead of mean or mode (ablation)IAC recency signal · ShareChatcache hit rateImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy