AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Stochastic Sparse Attention for Memory-Bound Inference

Kyle Lee, Corentin Delacour, Kevin Callahan-Coray, Kyle Jiang, Can Yaras, Samet Oymak, Tathagata Srimani, Kerem Yunus Camsari

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 stratified sampling instead of iid sampling (ablation)SANTA (S=16) · GSM8Kaccuracy (%): 5.51 -> 39.12Improved
use systematic instead of independent stratified sampling (ablation)S2ANTA · GSM8K (S=16)accuracy (%): 39.12 -> 44.63Improved
increase sample budget S from 2 to 256 (ablation)S2ANTA-sys · GSM8Kaccuracy (%): 1.11 -> 77.56Improved
replace iid sampling with systematic sampling at S=32 (ablation)SANTA · MMLUaccuracy (%): 33.81 -> 45.48Improved
use speculative per-tile flash allocation with S=256 vs S=2048 (ablation)S2ANTA-flash · RULER FWE (32k)accuracy (%): 66.20 -> 94.13Improved
replace S2ANTA-flash with S2ANTA-prop proportional allocationS2ANTA kernel · RULER (32k)smallest S matching SDPA: S=2048 -> S=128Improved
swap FlashAttention-2 decode for S2ANTA-flashLlama-3.1-8B-Instruct · end-to-end decodedecode speedup: 1.000x -> 1.250xImproved
use Bernoulli sampling with B=4 instead of B=1 (ablation)qK^T estimator · GSM8K (BitNet 2B)accuracy (%): 1.9 -> 64.5Improved
use mean group query instead of per-head query (ablation)Bernoulli qK^T (B=4) · GSM8K (BitNet 2B)K access (%) and accuracy: 97.9 -> 84.7 group access; 64.5 -> 63.8 accuracyMixed
increase SANTA budget S from 8 to 64 with Bernoulli B=4 (ablation)Bernoulli qK^T + SANTA · GSM8K (BitNet 2B)accuracy (%): 55.0 -> 63.8Improved
increase Bernoulli samples from B=4 to B=8 (ablation)Bernoulli qK^T · RULER FWE (4k)accuracy (%): 70.4 -> 89.4Improved
replace top-k with S2ANTA-sys at matched budget 16attention value stage · GSM8Kaccuracy (%): 62.88 -> 44.63Hurt

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