AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

IndexMem: Learned KV-Cache Eviction with Latent Memory for Long-Context LLM Inference

Xintong Yang, Hao Gu, Binxing Xu, Lujun Li, Bei Liu, Jiacheng Liu, Qiyuan Zhu, Yike Guo, Sirui Han

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
add learned indexer plus latent memory instead of heuristic eviction scoresKV eviction on Qwen3-8B · RULER 4K, 90% evictionaccuracy: 37.9 -> 46.2 vs KeyDiffImproved
use IndexMem rather than SnapKV at high evictionMistral-7B-v0.3 · RULER 16K, 90% evictionaccuracy: 21.3 -> 52.9Improved
IndexMem vs TOVA on a long-context setting with heavy evictionLlama-3.1-8B-Instruct · RULER 16K, 90% evictionaccuracy: 59.2 -> 56.0Hurt
IndexMem vs TOVA at 50% evictionMistral-7B-v0.3 · RULER 16K, 50% evictionaccuracy: 70.1 -> 68.5Hurt
add latent memory on top of indexer-only (ablation)IndexMem · LongBench, Llama-3.1-8B-InstructscoreImproved
add latent memory to indexer-only for NIAH (ablation)IndexMem · NIAH, 50% compressionretrieval robustnessImproved
plug latent memory into SnapKV (SnapKV+Memory) (ablation)SnapKV · LongBench tasksscoreImproved
increase decoding KV budget Bmax (ablation)all methods · AIME25 and Math500, Qwen3-8BaccuracyImproved
moderate eviction (about 50%) on LongBench triviaqa (ablation)IndexMem · LongBench triviaqascoreImproved
raise compression to 90% (ablation)IndexMem · LongBench trecscoreHurt

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