AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

QuoKA: Query-Oriented KV Selection for Efficient LLM Prefill

Dalton Jones, Junyoung Park, Matthew J Morse, Mingu Lee, Matthew Harper Langston, Christopher Lott

ICLR 2026 · Architectures and efficiency › Attention and transformers · approach: KV cache eviction and compression · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use QuoKA instead of SampleAttention for KV selection in chunked prefillLlama-3.2-3B-Instruct · RULER 32k, B_SA=1024score: 31.73 -> 57.01Improved
use QuoKA instead of SampleAttentionQwen-3-4B · RULER 32k, B_SA=1024score: 40.72 -> 74.83Improved
use QuoKA instead of SampleAttention on GPT-OSS-20B at 4kGPT-OSS-20B · RULER 4k, B_SA=1024score: 76.20 -> 78.92No clear effect
use QuoKA instead of SampleAttention for LongBench with small budgetLlama3.2-3B-Instruct · LongBench, B_SA=512normalized accuracy: 0.738 -> 0.945Improved
use QuoKA with 25% KV budget instead of full attentionQwen2.5-3B · RULER 32768score: 71.69 -> 68.84Hurt
use QuoKA with 25% KV budget instead of full attentionQwen3-4B · RULER 32768score: 88.54 -> 87.87No clear effect
score keys by cosine similarity instead of dot product (ablation)QuoKA · RULERsubselection quality: more than 10%Improved
reduce query subselection to N_Q = B_CP/16 (ablation)QuoKA · LongBench and RULERaccuracy: about 3% dropHurt

Later papers that compared with this method

1 later accepted paper reports QUOKA as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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