Saved in:
Bibliographic Details
Main Authors: Wei, Xiuying, Gulcehre, Caglar
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.28640
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914608651960320
author Wei, Xiuying
Gulcehre, Caglar
author_facet Wei, Xiuying
Gulcehre, Caglar
contents Efficient inference is critical for long-context language models, where attention computation and KV-cache access dominate the cost. Recent work RAT+, introduces a recurrence-augmented attention backbone that enables flexible dilated attention at inference time. In this paper, we investigate whether this exponentially decaying memory can also improve existing query-aware sparse inference methods. Using representative methods including Quest, MoBA, and SnapKV, we show that RAT+ consistently improves accuracy over standard attention across sparse budgets on eight needle-in-a-haystack tasks. We validate these gains both on the released checkpoints from the RAT+ paper and on OLMo2-7B, which we continue pretraining with the added memory module for 10B tokens. Finally, we propose two hypotheses explaining why this memory module benefits query-aware sparse inference and design targeted experiments to support them.
format Preprint
id arxiv_https___arxiv_org_abs_2605_28640
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Augmenting Attention with Exponentially Decaying Memory Improves Query-Aware KV Sparsity
Wei, Xiuying
Gulcehre, Caglar
Machine Learning
Efficient inference is critical for long-context language models, where attention computation and KV-cache access dominate the cost. Recent work RAT+, introduces a recurrence-augmented attention backbone that enables flexible dilated attention at inference time. In this paper, we investigate whether this exponentially decaying memory can also improve existing query-aware sparse inference methods. Using representative methods including Quest, MoBA, and SnapKV, we show that RAT+ consistently improves accuracy over standard attention across sparse budgets on eight needle-in-a-haystack tasks. We validate these gains both on the released checkpoints from the RAT+ paper and on OLMo2-7B, which we continue pretraining with the added memory module for 10B tokens. Finally, we propose two hypotheses explaining why this memory module benefits query-aware sparse inference and design targeted experiments to support them.
title Augmenting Attention with Exponentially Decaying Memory Improves Query-Aware KV Sparsity
topic Machine Learning
url https://arxiv.org/abs/2605.28640