AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

OBCache: Optimal Brain KV Cache Pruning for Efficient Long-Context LLM Inference

Yuzhe Gu, Xiyu Liang, Jiaojiao Zhao, Enmao Diao

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: KV cache eviction and compression · 0 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
replace attention score with OBCache-V&KH2O · RULER-4K query-agnosticavg accuracy: 31.7 -> 40.0Improved
replace attention score with OBCache-V&KAdaKV · RULER-4K query-agnosticavg accuracy: 43.0 -> 55.2Improved
replace attention score with OBCache-V&KAdaKV · RULER-4K query-awareavg accuracy: 75.7 -> 81.9Improved
replace attention score with OBCache-V&KH2O · RULER-4K query-awareavg accuracy: 57.5 -> 67.8Improved
replace attention score with OBCache-V&KTOVA · RULER-4K query-awareavg accuracy: 74.5 -> 76.7Improved
replace attention score with OBCache-V&KH2O · RULER-32K query-agnosticavg accuracy: 31.5 -> 36.7Improved
replace attention score with OBCache-V&KAdaKV · RULER-32K query-agnosticavg accuracy: 45.5 -> 55.1Improved
replace attention score with OBCache-VSnapKV · RULER-4K query-awareavg accuracy: 72.4 -> 71.9Hurt
replace attention score with OBCache-VSnapKV · RULER-4K query-agnosticavg accuracy: 37.9 -> 35.8Hurt
use key-pruning score instead of value-only score (ablation)AdaKV + OBCache · RULER-4K query-agnosticavg accuracy: 51.4 -> 55.0Improved
add key score to value score (V&K versus K only) (ablation)AdaKV + OBCache · RULER-4K query-agnosticavg accuracy: 55.0 -> 55.2No clear effect
use OBCache-K scoresAdaKV · LongBench 10% budgetscore: +1.2 query-aware; +2.6 query-agnosticImproved

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