@inproceedings{lee-etal-2026-score,
title = "{SCORE}: Similarity-Aware Contextual Overlap-Redundancy Eviction for Efficient {KV} Cache Compression in {LLM}s",
author = "Lee, Gilha and
Lee, Seungil and
Kim, Hyun",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2155/",
pages = "49020--49038",
ISBN = "979-8-89176-459-0",
abstract = "Recent advances in large language models (LLMs) have unlocked remarkable long-context capabilities, enabling breakthroughs across diverse NLP tasks. However, despite architectural progress, the key-value (KV) cache remains a critical memory bottleneck during inference. Prior work has explored cache optimization via eviction strategies, yet most rely on heuristic or single-axis importance metrics, neglecting the nuanced and dynamic interplay between layers and attention heads. In this paper, we propose SCORE (Similarity-aware Contextual Overlap-Redundancy Eviction), a novel framework that introduces a distance-based multi-level similarity metric to quantify and eliminate structural redundancy within the KV cache. By dynamically reallocating cache budgets across layers and heads and employing a redundancy-aware greedy token selection mechanism, SCORE preserves semantic diversity while minimizing memory overhead. Extensive experiments on long-context benchmarks such as LongBench and NeedleBench show that SCORE retains 96.2{\%} of full KV cache performance using only 1.5{\%} of the cache, consistently outperforming state-of-the-art baselines under strict memory constraints. These results underscore the value of fine-grained, context-aware cache management for scalable and efficient long-context inference in LLMs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lee-etal-2026-score">
<titleInfo>
<title>SCORE: Similarity-Aware Contextual Overlap-Redundancy Eviction for Efficient KV Cache Compression in LLMs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Gilha</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seungil</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hyun</namePart>
<namePart type="family">Kim</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Recent advances in large language models (LLMs) have unlocked remarkable long-context capabilities, enabling breakthroughs across diverse NLP tasks. However, despite architectural progress, the key-value (KV) cache remains a critical memory bottleneck during inference. Prior work has explored cache optimization via eviction strategies, yet most rely on heuristic or single-axis importance metrics, neglecting the nuanced and dynamic interplay between layers and attention heads. In this paper, we propose SCORE (Similarity-aware Contextual Overlap-Redundancy Eviction), a novel framework that introduces a distance-based multi-level similarity metric to quantify and eliminate structural redundancy within the KV cache. By dynamically reallocating cache budgets across layers and heads and employing a redundancy-aware greedy token selection mechanism, SCORE preserves semantic diversity while minimizing memory overhead. Extensive experiments on long-context benchmarks such as LongBench and NeedleBench show that SCORE retains 96.2% of full KV cache performance using only 1.5% of the cache, consistently outperforming state-of-the-art baselines under strict memory constraints. These results underscore the value of fine-grained, context-aware cache management for scalable and efficient long-context inference in LLMs.</abstract>
<identifier type="citekey">lee-etal-2026-score</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2155/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>49020</start>
<end>49038</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SCORE: Similarity-Aware Contextual Overlap-Redundancy Eviction for Efficient KV Cache Compression in LLMs
%A Lee, Gilha
%A Lee, Seungil
%A Kim, Hyun
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F lee-etal-2026-score
%X Recent advances in large language models (LLMs) have unlocked remarkable long-context capabilities, enabling breakthroughs across diverse NLP tasks. However, despite architectural progress, the key-value (KV) cache remains a critical memory bottleneck during inference. Prior work has explored cache optimization via eviction strategies, yet most rely on heuristic or single-axis importance metrics, neglecting the nuanced and dynamic interplay between layers and attention heads. In this paper, we propose SCORE (Similarity-aware Contextual Overlap-Redundancy Eviction), a novel framework that introduces a distance-based multi-level similarity metric to quantify and eliminate structural redundancy within the KV cache. By dynamically reallocating cache budgets across layers and heads and employing a redundancy-aware greedy token selection mechanism, SCORE preserves semantic diversity while minimizing memory overhead. Extensive experiments on long-context benchmarks such as LongBench and NeedleBench show that SCORE retains 96.2% of full KV cache performance using only 1.5% of the cache, consistently outperforming state-of-the-art baselines under strict memory constraints. These results underscore the value of fine-grained, context-aware cache management for scalable and efficient long-context inference in LLMs.
%U https://aclanthology.org/2026.emnlp-main.2155/
%P 49020-49038
Markdown (Informal)
[SCORE: Similarity-Aware Contextual Overlap-Redundancy Eviction for Efficient KV Cache Compression in LLMs](https://aclanthology.org/2026.emnlp-main.2155/) (Lee et al., EMNLP 2026)
ACL