AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Sparse Attention Across Multiple-Context KV Cache

Ziyi Cao, Qingyi Si, Jingbin Zhang, Bingquan Liu

AAAI 2026 · Architectures and efficiency › Attention and transformers · approach: KV cache eviction and compression · 6 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
add personalized bias to query vector (no recompute) (ablation)SamKV (Llama 3.1 8B) · Table 4 avg F1avg F1: 15.00 -> 15.40Improved
add personalized bias (with recompute) (ablation)SamKV (Llama 3.1 8B) · Table 4 avg F1avg F1: 22.21 -> 22.75Improved
add recomputation of sparse tokens (ablation)SamKV (Llama 3.1 8B) · Table 4 avg F1avg F1: 15.40 -> 22.75Improved
add middle-section KV selection (no recompute) (ablation)SamKV (Llama 3.1 8B) · Table 4 avg F1avg F1: 14.68 -> 15.00No clear effect
use KV fusion instead of overwriteSamKV (Mistral 7B) · 2WikiMQAF1: 27.04 -> 27.88Improved
use KV fusion instead of overwriteSamKV (Llama 3.1 8B) · HotpotQAF1: 35.27 -> 30.52Hurt
use SamKV (15% sequence) instead of full recomputationLlama 3.1 8B · HotpotQAF1: 24.12 -> 35.27Improved
use SamKV instead of full recomputationMistral 7B · MuSiQueF1: 19.94 -> 18.54Hurt
use SamKV instead of CacheBlend/EPICMistral 7B · 2WikiMQAF1: 20.42 / 19.40 -> 27.88Improved

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