Signals 4 · free daily AI digest

SANTA++: Sampling Attention through Representative Keys

Paper recorded by Signals 4 on 2026-09-28 in cs.CL. Abstract reproduced from arXiv; link to the original below.

Published 2026-09-28 on arXiv · recorded by Signals 4 on 2026-09-29

Category: cs.CL · 自然语言处理 · first seen 2026-09-29

Abstract

Attention often concentrates on a small subset of tokens in the context, but which subset matters changes from one query to the next. To exploit this changing structure, we introduce SANTA++, a training-free stochastic attention method that uses representative keys for memory-efficient selection without scanning the entire key-value (KV) cache. Cached keys are organized into teams, and the query s

Read on arXiv →

#10 most recent of 279 cs.CL papers we have recorded · ↑ newer: Which the Eye Fears: Writing with Read-Blindness Explains Massive Acti · ↓ older: Can LLMs Value the Right Evidence? Evidence-Value Misalignment in Dyna
Cite this page: SANTA++: Sampling Attention through Representative Keys: the #10 most recent of 279 cs.CL papers we have recorded (as of 2026-09-28). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/papers/santa-sampling-attention-through-representative-keys.html
Free to quote with attribution to “Signals 4 (Signals API)”. Machine-readable: papers.json
Related: More cs.CL papers · arXiv signals · All papers · Today in AI
Get 4 AI signals a day by email — free.
Subscribe free → See all plans →
Get 4 AI signals a day by email — free
All models · All repos · By company · Daily editions