---
title: "Adapting Sleep Foundation Models for Dense Sleep Spindle Localization"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2611.06546v1"
permalink: "https://feed.astralor.com/items/itm_d84c55ae8dc08930ab9e757b"
published: "2026-11-04T14:12:33.000Z"
score: 57
channel: "ai"
tags: ["推理效率","Adapting Sleep Foundation","Models","Dense Sleep Spindle","Localization Adapting","EEG","Adapting","SleepFM"]
---

# Adapting Sleep Foundation Models for Dense Sleep Spindle Localization

## At a glance

Adapting sleep foundation models to dense electroencephalography (EEG) event localization is challenging because multi-second representations may not preserve sufficient temporal detail for precise boundary recovery. We…

## Key points

- Adapting sleep foundation models to dense electroencephalography (EEG) event localization is challenging because multi-second representations may not preserve sufficient temporal detail for precise boundary recovery.
- We address sleep spindle localization by adapting a frozen SleepFM encoder with a local time-frequency pathway that preserves spindle-relevant spectral content and intra-token position.
- The local representation drives localization, while residual cross-attention incorporates SleepFM temporal context.

## Full text

Adapting sleep foundation models to dense electroencephalography (EEG) event localization is challenging because multi-second representations may not preserve sufficient temporal detail for precise boundary recovery. We address sleep spindle localization by adapting a frozen SleepFM encoder with a local time-frequency pathway that preserves spindle-relevant spectral content and intra-token position. The local representation drives localization, while residual cross-attention incorporates SleepFM temporal context. Dense predictions are obtained through localized token-to-sample reconstruction and temporal refinement. Tested on the Massive Online Data Annotation (MODA), the proposed model achieves an event-level $F_1$ of $0.8080 \pm 0.0040$ at IoU 0.2 across six folds. Ablations show that localization primarily depends on local spectral and sub-token temporal information, while the pretrained representation provides complementary context. These results indicate that dense adaptation benefits from preserving information below the pretrained token scale. Code will be released upon acceptance.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_d84c55ae8dc08930ab9e757b