@inproceedings{wang-etal-2026-take,
title = "{TAKE}: Task-Aware Chunked {KV} Cache Eviction for Efficient Long-Context {LLM} Prefill",
author = "Wang, Rui and
Jia, Nan and
Li, Jiahui and
Yang, Qingyi and
Hao, Yixue and
Hu, Long",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.260/",
pages = "5449--5463",
ISBN = "979-8-89176-462-0",
abstract = "Memory bottlenecks severely constrain the deployment of large language models (LLMs) in long-context scenarios. While existing global pruning methods reduce memory usage during the decoding phase, they largely ignore out-of-memory errors in the prefill stage before generation begins. To address this, we present Task-Aware Chunked KV Cache Eviction (TAKE), a training-free framework for memory-efficient long-context prefill. TAKE refines the prefill process by partitioning long sequences into chunks and performing incremental KV eviction, thereby avoiding the computational overhead of full-sequence processing. Crucially, to mitigate the semantic discontinuity introduced by chunk-wise processing, we introduce accumulated task-aware probe tokens and a delayed eviction strategy to reduce information loss. Extensive evaluations show that TAKE achieves the highest efficiency, slashing peak memory usage to 59.53{\%} of the baseline while accelerating Time-To-First-Token (TTFT) by over 60{\%} under 128k-token inference. Our code is available at [Source Code of TAKE](https://github.com/ObitoJN/TAKE)."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2026-take">
<titleInfo>
<title>TAKE: Task-Aware Chunked KV Cache Eviction for Efficient Long-Context LLM Prefill</title>
</titleInfo>
<name type="personal">
<namePart type="given">Rui</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nan</namePart>
<namePart type="family">Jia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiahui</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qingyi</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yixue</namePart>
<namePart type="family">Hao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Long</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Memory bottlenecks severely constrain the deployment of large language models (LLMs) in long-context scenarios. While existing global pruning methods reduce memory usage during the decoding phase, they largely ignore out-of-memory errors in the prefill stage before generation begins. To address this, we present Task-Aware Chunked KV Cache Eviction (TAKE), a training-free framework for memory-efficient long-context prefill. TAKE refines the prefill process by partitioning long sequences into chunks and performing incremental KV eviction, thereby avoiding the computational overhead of full-sequence processing. Crucially, to mitigate the semantic discontinuity introduced by chunk-wise processing, we introduce accumulated task-aware probe tokens and a delayed eviction strategy to reduce information loss. Extensive evaluations show that TAKE achieves the highest efficiency, slashing peak memory usage to 59.53% of the baseline while accelerating Time-To-First-Token (TTFT) by over 60% under 128k-token inference. Our code is available at [Source Code of TAKE](https://github.com/ObitoJN/TAKE).</abstract>
<identifier type="citekey">wang-etal-2026-take</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.260/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>5449</start>
<end>5463</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T TAKE: Task-Aware Chunked KV Cache Eviction for Efficient Long-Context LLM Prefill
%A Wang, Rui
%A Jia, Nan
%A Li, Jiahui
%A Yang, Qingyi
%A Hao, Yixue
%A Hu, Long
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F wang-etal-2026-take
%X Memory bottlenecks severely constrain the deployment of large language models (LLMs) in long-context scenarios. While existing global pruning methods reduce memory usage during the decoding phase, they largely ignore out-of-memory errors in the prefill stage before generation begins. To address this, we present Task-Aware Chunked KV Cache Eviction (TAKE), a training-free framework for memory-efficient long-context prefill. TAKE refines the prefill process by partitioning long sequences into chunks and performing incremental KV eviction, thereby avoiding the computational overhead of full-sequence processing. Crucially, to mitigate the semantic discontinuity introduced by chunk-wise processing, we introduce accumulated task-aware probe tokens and a delayed eviction strategy to reduce information loss. Extensive evaluations show that TAKE achieves the highest efficiency, slashing peak memory usage to 59.53% of the baseline while accelerating Time-To-First-Token (TTFT) by over 60% under 128k-token inference. Our code is available at [Source Code of TAKE](https://github.com/ObitoJN/TAKE).
%U https://aclanthology.org/2026.findings-emnlp.260/
%P 5449-5463
Markdown (Informal)
[TAKE: Task-Aware Chunked KV Cache Eviction for Efficient Long-Context LLM Prefill](https://aclanthology.org/2026.findings-emnlp.260/) (Wang et al., Findings 2026)
ACL