# sleep

Published articles for sleep.

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## Hypnos: Learning Generalisable Representations of Sleep Physiology via Next-Token Prediction

DevFeed: [Hypnos: Learning Generalisable Representations of Sleep Physiology via Next-Token Prediction](<https://devfeed.tech/articles/hypnos-learning-generalisable-representations-of-sleep-physiology-via-next-token-prediction-120047.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2606.09605>)

Author: Jonathan F. Carter, Lionel Tarassenko

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [multimodal-ai](<https://devfeed.tech/topics/multimodal-ai.md>), [Language models](<https://devfeed.tech/topics/language-models.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>)

Tags: [atrial-fibrillation](<https://devfeed.tech/tags/atrial-fibrillation.md>), [classification](<https://devfeed.tech/tags/classification.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [ecg](<https://devfeed.tech/tags/ecg.md>), [eeg](<https://devfeed.tech/tags/eeg.md>), [embeddings](<https://devfeed.tech/tags/embeddings.md>), [foundation-model](<https://devfeed.tech/tags/foundation-model.md>), [next-token-prediction](<https://devfeed.tech/tags/next-token-prediction.md>), [quantization](<https://devfeed.tech/tags/quantization.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [training](<https://devfeed.tech/tags/training.md>), [transformer](<https://devfeed.tech/tags/transformer.md>), [using](<https://devfeed.tech/tags/using.md>)

### AI overview

Hypnos learns representations from eight physiological sensing modalities in more than 20,000 overnight polysomnography recordings. It converts each modality into discrete tokens with residual vector quantization, then trains an autoregressive RQ-Transformer to predict the modalities' next tokens jointly. The resulting model can generate downstream-task embeddings from continuous data using any subset of its supported modalities. The abstract reports that Hypnos outperforms existing foundation models on most evaluated benchmarks, matches strong supervised baselines for sleep-stage classification with 100 times less labeled data, and exceeds a dedicated ECG foundation model on atrial-fibrillation detection. These are reported results; the excerpt does not provide benchmark details.

### Source excerpt

arXiv:2606.09605v2 Announce Type: replace Abstract: Foundation models offer a promising route to compress multi-modal physiological signals into compact representations of human health, with broad applications across sleep medicine, cardiology, neurology and other healthcare domains. Existing models have typically been trained with masked-reconstruction or contrastive objectives. However, masked reconstruction may be poorly suited to the stochastic nature of these signals, while contrastive approaches rely on positive-pair definitions despite the semantic invariances of physiological signals being poorly understood. In this work, we show that next-token prediction is a simple and scalable alternative. We develop Hypnos, a multi-modal sleep foundation model trained using eight different sensing modalities (e.g. EEG, ECG, respiratory signals) drawn from over 20{,}000 overnight polysomnography recordings. We tokenize each modality into streams of discrete tokens using residual vector quantization, then train a large auto-regressive RQ-Transformer to jointly predict the next token across all modalities in parallel. After training, Hypnos can be applied to continuous streams of sensor data from any subset of supported modalities, generating embeddings for downstream tasks. Hypnos significantly outperforms existing foundation models on most benchmarks evaluated. In sleep stage classification, we match the performance of strong supervised baselines on held-out test sets whilst using 100\times less labelled data. Hypnos even generalises beyond sleep recordings, surpassing a dedicated ECG foundation model at detecting atrial fibrillation. Our results demonstrate that auto-regressive token modelling is a strong self-supervised approach for representation learning from multi-modal physiological signals. Code and model weights: https://github.com/joncarter1/hypnos

## Backbone-Informed Adaptation after EEG Pretraining: A Capacity-Controlled Study

DevFeed: [Backbone-Informed Adaptation after EEG Pretraining: A Capacity-Controlled Study](<https://devfeed.tech/articles/backbone-informed-adaptation-after-eeg-pretraining-a-capacity-controlled-study-120571.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.07192>)

Author: Mingyang Jiang, Yamin Li, Fan Ma, Daniel Moyer, Hua Xu, Catie Chang

Published: 2026-11-06T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [mlp](<https://devfeed.tech/topics/mlp.md>)

Tags: [adapter](<https://devfeed.tech/tags/adapter.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [comparison](<https://devfeed.tech/tags/comparison.md>), [control](<https://devfeed.tech/tags/control.md>), [counts](<https://devfeed.tech/tags/counts.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [eeg](<https://devfeed.tech/tags/eeg.md>), [emotion](<https://devfeed.tech/tags/emotion.md>), [event](<https://devfeed.tech/tags/event.md>), [performance](<https://devfeed.tech/tags/performance.md>), [pretraining](<https://devfeed.tech/tags/pretraining.md>), [protocol](<https://devfeed.tech/tags/protocol.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [tasks](<https://devfeed.tech/tags/tasks.md>), [token](<https://devfeed.tech/tags/token.md>)

### AI overview

With trainable parameters approximately matched and the pretrained encoder frozen, a backbone-informed EEG adapter showed no consistent advantage over a token-wise multilayer perceptron that transforms feature vectors independently. Across two EEG foundation models and emotion recognition, sleep staging, and EEG event classification, results varied by backbone and training protocol: the token-wise control performed better on two tasks for one backbone, while the native adapter's sleep-staging advantage narrowed when the sequence-processing component was also trained. The abstract supports evaluating such adapters against capacity-matched alternatives; it does not provide full methods or numerical results.

### Source excerpt

arXiv:2611.07192v1 Announce Type: new Abstract: Electroencephalography (EEG) foundation models learn representations that combine information across channels and time. This structure motivates adapters that model the same relationships during downstream learning. However, it is not clear whether such a design outperforms other candidate adapters. Moreover, in order to arrive at a fair comparison between candidate adapters, we also need to consider their trainable parameters. Here, we ask whether a backbone-informed (native) adapter makes better use of a given capacity by comparing it against a token-wise multilayer perceptron control that transforms each feature vector independently. We approximately match their trainable parameter counts and keep the pretrained encoder fixed. Across two EEG foundation models and tasks covering emotion recognition, sleep staging, and EEG event classification, the native adapter shows no consistent advantage. For one backbone, differences between the designs are small relative to variation across runs. For the other, the token-wise control performs better on emotion recognition and EEG event classification. On sleep staging, the native adapter's advantage over the token-wise control is reduced when we also train the component that processes sequences of EEG segments. The relative performance of the two adapters depends on the backbone, task, and downstream training protocol. These findings motivate evaluating backbone-informed adapters against capacity-matched alternatives.

## Do agents need to sleep?

DevFeed: [Do agents need to sleep?](<https://devfeed.tech/articles/do-agents-need-to-sleep-118301.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.02216>)

Author: Ido Haber, Giulio Tononi

Published: 2026-11-05T05:00:00Z

Content type: opinion

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [long-running](<https://devfeed.tech/topics/long-running.md>), [AI Context Management](<https://devfeed.tech/topics/ai-context-management.md>), [llamaindex](<https://devfeed.tech/topics/llamaindex.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [continuous](<https://devfeed.tech/tags/continuous.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [digital](<https://devfeed.tech/tags/digital.md>), [engineers](<https://devfeed.tech/tags/engineers.md>), [environment](<https://devfeed.tech/tags/environment.md>), [errors](<https://devfeed.tech/tags/errors.md>), [interaction](<https://devfeed.tech/tags/interaction.md>), [learning](<https://devfeed.tech/tags/learning.md>), [llm](<https://devfeed.tech/tags/llm.md>), [memory](<https://devfeed.tech/tags/memory.md>), [plasticity](<https://devfeed.tech/tags/plasticity.md>), [precision](<https://devfeed.tech/tags/precision.md>), [prediction](<https://devfeed.tech/tags/prediction.md>), [problems](<https://devfeed.tech/tags/problems.md>), [q-bio-nc](<https://devfeed.tech/tags/q-bio-nc.md>), [retrieval](<https://devfeed.tech/tags/retrieval.md>), [semantic](<https://devfeed.tech/tags/semantic.md>), [semantic-drift](<https://devfeed.tech/tags/semantic-drift.md>), [sessions](<https://devfeed.tech/tags/sessions.md>), [sleep](<https://devfeed.tech/tags/sleep.md>)

### AI overview

The paper argues that persistent LLM agents face memory problems associated with unchecked plasticity: semantic drift, poorer retrieval, stored errors, and capacity saturation. It maps the three phases of the selectionist "ringing, resonance and renormalization" account of sleep to operations feasible in digital memory architectures. From this analogy, it predicts that selectionist memory consolidation will outperform instructionist alternatives as memory stores grow large. This is presented as a falsifiable prediction; the supplied abstract does not report comparative test results.

### Source excerpt

arXiv:2611.02216v2 Announce Type: cross Abstract: LLM-based agents, which increasingly operate persistently across sessions and accumulate memory through interaction, already exhibit the familiar symptoms of unchecked plasticity: semantic drift, degraded retrieval precision, propagation of stored errors, and saturation of memory capacity. Drawing on the selectionist "ringing, resonance and renormalization" (RRR) account of sleep, we propose that each of its three phases maps onto an operation feasible in digital memory architectures. Engineers building stateful agents, meanwhile, are independently rediscovering the problems sleep evolved to solve: what to forget, and when to do the forgetting. From this we derive the falsifiable prediction that selectionist consolidation will outperform instructionist alternatives as memory stores grow large. The core claim concerns the nature of continuous learning itself: plasticity has a price, and for any system that learns continuously under finite capacity while coupled to its environment, that price is sleep.

## Sleep-Disordered Breathing Detection with Quantum Recurrent Neural Networks

DevFeed: [Sleep-Disordered Breathing Detection with Quantum Recurrent Neural Networks](<https://devfeed.tech/articles/sleep-disordered-breathing-detection-with-quantum-recurrent-neural-networks-116987.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.04818>)

Author: Alejandro Mayorga-Redondo, Jos\\'e Daniel Viqueira, Diego Alvarez-Estevez, Eduardo Mosqueira-Rey, Andr\\'es G\\'omez

Published: 2026-11-04T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Quantum Computing](<https://devfeed.tech/topics/quantum-computing.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>)

Tags: [classification](<https://devfeed.tech/tags/classification.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [quant-ph](<https://devfeed.tech/tags/quant-ph.md>), [quantum](<https://devfeed.tech/tags/quantum.md>), [quantum-computing](<https://devfeed.tech/tags/quantum-computing.md>), [recurrent-neural-networks](<https://devfeed.tech/tags/recurrent-neural-networks.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [time-series](<https://devfeed.tech/tags/time-series.md>)

### AI overview

A variational classical-quantum recurrent model classified apneic versus normal breathing episodes and showed performance comparable to classical recurrent neural networks across multiple real-world datasets. It performed better on the smallest dataset but slightly worse in the most complex setting. The authors report competitive results with far fewer parameters and few qubits, along with good generalization, while noting current hardware and algorithmic limitations. The excerpt does not provide specific metrics or implementation details.

### Source excerpt

arXiv:2611.04818v1 Announce Type: new Abstract: This work investigates the application of Quantum Recurrent Neural Networks to the binary classification of sleep-disordered breathing events in time-series data, aiming to distinguish between apneic and normal breathing episodes. A variational classical--quantum architecture is proposed and evaluated against classical Recurrent Neural Network models. Experiments were conducted on multiple real-world datasets with varying levels of complexity to assess the strengths and limitations of the quantum approach. In the comparative analysis, both classical and quantum models showed comparable performance. The quantum model achieved superior performance on the smallest dataset, while yielding slightly weaker results in the most complex setting. However, the quantum approach exhibited notable strengths, such as competitive performance despite a drastically reduced parameter count and low-qubit regime, as well as good generalization capabilities. These results suggest that, despite current hardware and algorithmic limitations, variational quantum recurrent models represent a promising alternative for time-series classification as Quantum Computing technologies continue to mature.

## Linux Mint is 'bringing back' hibernation to laptops

DevFeed: [Linux Mint is 'bringing back' hibernation to laptops](<https://devfeed.tech/articles/linux-mint-is-bringing-back-hibernation-to-laptops-112915.md>)

Original publisher: [Read original article](<https://www.omgubuntu.co.uk/2026/11/linux-mint-hibernation-sleep-mode>)

Author: Joey Sneddon

Published: 2026-11-03T00:00:03Z

Content type: news

Language: en

Sources: [OMG! Ubuntu](<https://devfeed.tech/sources/omg-ubuntu.md>)

Topics: [Linux Mint](<https://devfeed.tech/topics/linuxmint.md>)

Tags: [battery](<https://devfeed.tech/tags/battery.md>), [hibernation](<https://devfeed.tech/tags/hibernation.md>), [laptops](<https://devfeed.tech/tags/laptops.md>), [linux](<https://devfeed.tech/tags/linux.md>), [linux-mint](<https://devfeed.tech/tags/linux-mint.md>), [news](<https://devfeed.tech/tags/news.md>), [power](<https://devfeed.tech/tags/power.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [suspend](<https://devfeed.tech/tags/suspend.md>), [system-administration](<https://devfeed.tech/tags/system-administration.md>)

### AI overview

Linux Mint 27 is planned to add a Sleep Modes page in System Administration for enabling hibernation and choosing sleep behavior. Unlike suspend, which uses power and can lose the session if the battery runs flat, hibernation saves the session to disk and shuts down; it resumes more slowly and needs substantial swap space. Hybrid Sleep both suspends and saves the session to disk, while Suspend-then-Hibernate suspends first and hibernates when the battery becomes critically low. Standard suspend will remain available. The excerpt says Ubuntu removed hibernation from its available sleep modes; it is off by default there and is usually blocked with Secure Boot active.

### Source excerpt

Linux Mint's next release will offer a hibernation mode, so you can close your laptop mid-session and not worry about returning to a drained battery. A new Sleep Modes page in the System Administration tool in Linux Mint 27, out this December, makes it easy to enable hibernation and ensure it's setup the way you need. Current versions of Linux Mint suspend when you close the laptop lid. Though snoozing, the device continues to use some power. If the battery runs out while sleeping, the suspended session is kaput - an outcome "worse than a normal shutdown", say Mint. Hibernation [...] You're reading Linux Mint is 'bringing back' hibernation to laptops, a blog post from OMG! Ubuntu. Do not reproduce elsewhere without permission.

## Build an AI Wellbeing App with Fireworks and Fitbit via Auth0 Token Vault.

DevFeed: [Build an AI Wellbeing App with Fireworks and Fitbit via Auth0 Token Vault.](<https://devfeed.tech/articles/build-an-ai-wellbeing-app-with-fireworks-and-fitbit-via-auth0-token-vault-114135.md>)

Original publisher: [Read original article](<https://auth0.com/blog/build-an-ai-wellbeing-app-with-fireworks-fitbit-via-auth0-token-vault/>)

Author: Deepu K Sasidharan

Published: 2026-11-02T19:25:00Z

Content type: tutorial

Language: en

Sources: [Auth0 Blog](<https://devfeed.tech/sources/auth0-blog.md>)

Topics: [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [Auth0](<https://devfeed.tech/topics/auth0.md>), [function calling](<https://devfeed.tech/topics/function-calling.md>), [Secure token management](<https://devfeed.tech/topics/secure-token-management.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-sdk](<https://devfeed.tech/tags/ai-sdk.md>), [app](<https://devfeed.tech/tags/app.md>), [auth0](<https://devfeed.tech/tags/auth0.md>), [build](<https://devfeed.tech/tags/build.md>), [daily](<https://devfeed.tech/tags/daily.md>), [fitbit](<https://devfeed.tech/tags/fitbit.md>), [next-js](<https://devfeed.tech/tags/next-js.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [token](<https://devfeed.tech/tags/token.md>), [vault](<https://devfeed.tech/tags/vault.md>), [wellbeing](<https://devfeed.tech/tags/wellbeing.md>)

### AI overview

Daily Wellbeing retrieves the signed-in user's Google Health readings through Auth0 Token Vault, builds a latest-completed-day report with seven-day personal averages, and streams Fireworks suggestions only after the user explicitly chooses to share the summary. The report preserves recorded local dates, distinguishes missing values from zero, counts available comparison days, and skips data streams without permission. Users can review readings without inference. Fireworks receives a normalized summary through a server-side tool; its prompt limits advice to general wellbeing and excludes diagnosis and treatment, though the source cautions that prompt rules do not guarantee suitable responses. Setup requires a personal Google account linked to a health profile, Google Cloud OAuth configuration, an Auth0 tenant with Token Vault and Connected Accounts, and a Fireworks key and compatible model. Google OAuth test-mode refresh tokens expire after seven days, requiring reconnection.

### Source excerpt

Build Daily Wellbeing, a Next.js app that connects Fitbit through Auth0 Token Vault, compares your daily activity and sleep with your averages, and uses Fireworks to suggest what to try today.

## Nedimo: An Evidence-Grounded Multilingual Pipeline for AI-Generated Sleep Stories

DevFeed: [Nedimo: An Evidence-Grounded Multilingual Pipeline for AI-Generated Sleep Stories](<https://devfeed.tech/articles/nedimo-an-evidence-grounded-multilingual-pipeline-for-ai-generated-sleep-stories-112000.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.40970>)

Author: Georgios Makridis, Stefanos Kampouris

Published: 2026-11-02T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Generative AI](<https://devfeed.tech/topics/generative-ai.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [generative-ai-tools](<https://devfeed.tech/topics/generative-ai-tools.md>), [Interactive AI Systems](<https://devfeed.tech/topics/interactive-ai-systems.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [audio](<https://devfeed.tech/tags/audio.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-cl](<https://devfeed.tech/tags/cs-cl.md>), [cs-hc](<https://devfeed.tech/tags/cs-hc.md>), [cs-sd](<https://devfeed.tech/tags/cs-sd.md>), [effects](<https://devfeed.tech/tags/effects.md>), [evaluation](<https://devfeed.tech/tags/evaluation.md>), [large-language-model](<https://devfeed.tech/tags/large-language-model.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [planning](<https://devfeed.tech/tags/planning.md>), [product-data](<https://devfeed.tech/tags/product-data.md>), [protocol](<https://devfeed.tech/tags/protocol.md>), [prototype](<https://devfeed.tech/tags/prototype.md>), [requirements](<https://devfeed.tech/tags/requirements.md>), [retry](<https://devfeed.tech/tags/retry.md>), [script](<https://devfeed.tech/tags/script.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [speech](<https://devfeed.tech/tags/speech.md>), [speech-synthesis](<https://devfeed.tech/tags/speech-synthesis.md>), [stories](<https://devfeed.tech/tags/stories.md>), [text-to-speech](<https://devfeed.tech/tags/text-to-speech.md>)

### AI overview

Nedimo presents a multilingual pipeline for generating low-arousal sleep stories, while treating evidence for bedtime audio as mixed and making no claim of clinical efficacy. Deterministic planning sets explicit content and delivery parameters for language-model generation; scripts follow a four-phase attention taper and undergo structural and lexical checks, a bounded error-aware retry, and quarantine if unresolved. Provider-agnostic text-to-speech is followed by paragraph-aware joining, two-pass loudness normalization, and a publication gate that re-measures the encoded file against fixed loudness, loudness-range, and true-peak limits. The data model records stop position, replay, abandonment, and completion as ambiguous behavioral proxies, not clinical outcomes. The authors propose technical evaluation criteria and a prospective protocol to test sleep-onset effects, including possible reliance.

### Source excerpt

arXiv:2610.40970v1 Announce Type: new Abstract: Generative audio systems can produce large libraries of personalized bedtime content, but ordinary media objectives (retention, completion, and novelty) are poorly aligned with the goal of supporting a transition toward sleep. Direct evidence for narrated bedtime audio in adults is mixed, and a recent active-controlled trial did not support such audio as a standalone treatment for sleep disturbance. We translate this uncertain evidence into conservative engineering requirements and present a multilingual pipeline for generating and narrating low-arousal sleep stories. The implemented prototype separates deterministic planning from large language model generation, represents content and delivery parameters explicitly, shapes each script through a four-phase attention taper, validates structural and lexical constraints, performs a bounded error-aware retry, and quarantines unresolved outputs. Provider-agnostic text-to-speech synthesis is followed by paragraph-aware joining, two-pass loudness normalization, and a publication gate that re-measures the saved, encoded file against fixed loudness, loudness-range, and true-peak limits. The product data model records stop position, replay, early abandonment, and completion while treating these events as ambiguous behavioral proxies rather than clinical outcomes. We contribute an evidence-to-requirement mapping, a traceable guardrail architecture with explicit acceptance thresholds, technical evaluation criteria, and a prospective protocol for testing sleep-onset effects. No clinical efficacy is claimed; the system is positioned as a relaxation and attentional-support tool whose benefit and possible reliance must be evaluated empirically.

## Monthly News - October 2026

DevFeed: [Monthly News - October 2026](<https://devfeed.tech/articles/monthly-news-october-2026-110213.md>)

Original publisher: [Read original article](<https://blog.linuxmint.com/?p=5093>)

Author: Clem

Published: 2026-10-31T13:32:51Z

Content type: news

Language: en

Sources: [Linux Mint](<https://devfeed.tech/sources/the-linux-mint-blog.md>)

Topics: [Linux Mint](<https://devfeed.tech/topics/linuxmint.md>), [Linux](<https://devfeed.tech/topics/linux.md>), [Tool](<https://devfeed.tech/topics/tool.md>), [User interface design](<https://devfeed.tech/topics/ui-design.md>), [Image](<https://devfeed.tech/topics/image.md>), [friendly interactive shell](<https://devfeed.tech/topics/fish.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [features](<https://devfeed.tech/tags/features.md>), [linux](<https://devfeed.tech/tags/linux.md>), [linux-mint](<https://devfeed.tech/tags/linux-mint.md>), [main-topics](<https://devfeed.tech/tags/main-topics.md>), [monthly](<https://devfeed.tech/tags/monthly.md>), [news](<https://devfeed.tech/tags/news.md>), [october](<https://devfeed.tech/tags/october.md>), [october-2026](<https://devfeed.tech/tags/october-2026.md>), [sleep](<https://devfeed.tech/tags/sleep.md>)

### AI overview

The next Linux Mint version is planned to add easier sleep-mode setup, new utilities, and a rewritten community site. A System Administration page will help configure Hibernation, which saves the session to disk without using power but is slower and requires substantial swap space; Hybrid Sleep and Suspend-then-Hibernate are also planned. Nemo's preview pane will show audio, image, and video previews and media details. Superlight will monitor system resources, processes, battery history, and storage use. The Velocitty terminal will offer themed, movable tabs, custom launchable tabs, and configurable shells and startup commands. Unfold will handle basic archive viewing, creation, and extraction using 7zip and 7zip-RAR, including supported encrypted formats. Xenity replaces Zenity in Cinnamon spices, and the Community Website now uses GitHub authentication and stores less sensitive data. These features are previews and remain in development.

### Source excerpt

Hello everyone, Before we start with the news, I'd like to thank all the people who support this project. Your help is greatly appreciated. Thank you. As always, please note that the features presented below are a preview. They're still being worked on and are planned for the next version of Linux Mint. Sleep Modes [...]

## MM-Net: A Multimodal Multi-Task Network for Sleep Staging and Sleep Apnea Assessment After Ischemic Stroke

DevFeed: [MM-Net: A Multimodal Multi-Task Network for Sleep Staging and Sleep Apnea Assessment After Ischemic Stroke](<https://devfeed.tech/articles/mm-net-a-multimodal-multi-task-network-for-sleep-staging-and-sleep-apnea-assessment-after-ischemic-stroke-103158.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.36583>)

Author: Md Wahiduzzaman Suva, Esm E Moula Chowdhury Abha, Md Imtiaj Alam Sajin, Md Iftekharul Mobin, Jia Uddin

Published: 2026-10-28T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Embeddings](<https://devfeed.tech/topics/embeddings.md>)

Tags: [assessment](<https://devfeed.tech/tags/assessment.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [eess-sp](<https://devfeed.tech/tags/eess-sp.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [staging](<https://devfeed.tech/tags/staging.md>)

### AI overview

MM-Net jointly predicts sleep stages and respiratory events after ischemic stroke, combining sleep-adapted EEG-EOG representations and engineered neural descriptors with learned cardiorespiratory features through cross-attention and shared temporal modeling. Cross-fitting and ridge-based embedding alignment address representation inconsistency between patients used to train downstream models and unseen evaluation patients. In ten-fold patient-independent evaluation on 99 iSLEEPS patients, it achieved 76.3% staging accuracy (κ=0.656) and 0.825 respiratory-event AUC, outperforming evaluated baselines under matched protocols. Alignment mattered for encoder adaptation: staging accuracy rose 1.4 percentage points over a frozen encoder with alignment (p=0.020), versus 0.2 points without it. On Sleep-EDF, staging accuracy was 86.4%, above reimplemented single-EEG architectures; on ISRUC-I, it reached 80.3% staging accuracy and 0.889 respiratory AUC, exceeding retrained staging baselines. Without retraining, respiratory AUC was 0.833 on ISRUC-I and 0.841 on UCD.

### Source excerpt

arXiv:2610.36583v1 Announce Type: cross Abstract: Automated polysomnography after ischemic stroke is challenging because of altered sleep-related neural activity and concurrent respiratory disturbances. Existing stroke-specific approaches mainly address sleep-stage classification, leaving joint analysis of sleep architecture and respiratory events underexplored. Moreover, adapting pretrained EEG encoders introduces a representation inconsistency when downstream models are trained on embeddings from previously seen patients but evaluated on unseen individuals. We propose MM-Net, a multimodal multi-task framework that resolves this inconsistency through cross-fitting and ridge-based embedding alignment. Sleep-adapted EEG-EOG representations and engineered neural descriptors are fused with learned cardiorespiratory features through cross-attention and shared temporal modeling for simultaneous sleep staging and respiratory-event detection. On iSLEEPS (99 patients with subacute ischemic stroke), ten-fold patient-independent evaluation achieved 76.3% staging accuracy (Cohen's $\kappa$ = 0.656) and 0.825 respiratory-event AUC, outperforming the evaluated baselines under matched protocols. The sleep-adapted encoder improved the downstream model when its embeddings were aligned, raising staging accuracy by 1.4 percentage points over the frozen encoder ($p$ = 0.020) versus 0.2 points without alignment. MM-Net also achieved 86.4% staging accuracy on Sleep-EDF, above every single-EEG architecture reimplemented on the same recordings, and 80.3% staging accuracy with 0.889 respiratory AUC on ISRUC-I, exceeding its retrained staging baselines. Without retraining, the iSLEEPS-trained model retained respiratory AUCs of 0.833 on ISRUC-I and 0.841 on UCD. These findings support representation-consistent neural adaptation and joint physiological modeling across stroke, healthy-sleep, and sleep-disorder cohorts.

## The sleep framework: A new way for LLMs to retain knowledge in continual learning

DevFeed: [The sleep framework: A new way for LLMs to retain knowledge in continual learning](<https://devfeed.tech/articles/the-sleep-framework-a-new-way-for-llms-to-retain-knowledge-in-continual-learning-103437.md>)

Original publisher: [Read original article](<https://research.google/blog/the-sleep-framework-a-new-way-for-llms-to-retain-knowledge-in-continual-learning/>)

Published: 2026-10-27T17:35:00Z

Content type: article

Language: en

Sources: [The latest research from Google](<https://devfeed.tech/sources/the-latest-research-from-google.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Large Language Model](<https://devfeed.tech/topics/llm.md>), [llm-reasoning](<https://devfeed.tech/topics/llm-reasoning.md>), [Training AI Models](<https://devfeed.tech/topics/training-ai-models.md>), [Mixture of Experts (MoE)](<https://devfeed.tech/topics/mixture-of-experts-moe.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [abstract](<https://devfeed.tech/tags/abstract.md>), [abstraction](<https://devfeed.tech/tags/abstraction.md>), [activation](<https://devfeed.tech/tags/activation.md>), [adapting](<https://devfeed.tech/tags/adapting.md>), [agent](<https://devfeed.tech/tags/agent.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-systems](<https://devfeed.tech/tags/ai-systems.md>), [allows](<https://devfeed.tech/tags/allows.md>), [also](<https://devfeed.tech/tags/also.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [automated](<https://devfeed.tech/tags/automated.md>), [autonomous](<https://devfeed.tech/tags/autonomous.md>), [continual-learning](<https://devfeed.tech/tags/continual-learning.md>), [education-innovation](<https://devfeed.tech/tags/education-innovation.md>), [framework](<https://devfeed.tech/tags/framework.md>), [general-science](<https://devfeed.tech/tags/general-science.md>), [knowledge](<https://devfeed.tech/tags/knowledge.md>), [learning](<https://devfeed.tech/tags/learning.md>), [llms](<https://devfeed.tech/tags/llms.md>), [machine-intelligence](<https://devfeed.tech/tags/machine-intelligence.md>), [natural-language-processing](<https://devfeed.tech/tags/natural-language-processing.md>), [sleep](<https://devfeed.tech/tags/sleep.md>)

### AI overview

The sleep framework alternates active interaction with offline consolidation so an LLM can transfer short-term context into long-term parameters. During sleep, pre-allocated Mixture-of-Experts memory blocks are updated using self-generated practice examples: consolidation uses on-policy distillation and semantic rewards to select useful knowledge, while a dreaming loop diversifies synthetic experiences and filters them by learning value. The authors report retaining long-context knowledge up to 10M tokens and gains across translation, factual learning, and mathematical reasoning. In sequential translation on unseen languages, it had near-zero performance drop while ICL, Cartridges, and supervised fine-tuning showed severe forgetting; on BABILong it scaled near-perfectly to 10M tokens and outperformed GPT-4 and Titans. These are reported results from the supplied excerpt; it does not provide enough methodological detail to assess them independently.

### Source excerpt

Education Innovation

## Week Notes 26#43

DevFeed: [Week Notes 26#43](<https://devfeed.tech/articles/week-notes-26-43-100138.md>)

Original publisher: [Read original article](<https://www.jvt.me/week-notes/2026/43/>)

Author: Jamie Tanna

Published: 2026-10-25T21:29:00Z

Content type: article

Language: en

Sources: [Jamie Tanna](<https://devfeed.tech/sources/articles-on-jamie-tanna-software-engineer.md>)

Topics: [cve remediation](<https://devfeed.tech/topics/cve-remediation.md>), [Developer experience](<https://devfeed.tech/topics/developer-experience.md>), [ai-experiments](<https://devfeed.tech/topics/ai-experiments.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [apt](<https://devfeed.tech/tags/apt.md>), [certificate](<https://devfeed.tech/tags/certificate.md>), [chainguard](<https://devfeed.tech/tags/chainguard.md>), [compromise](<https://devfeed.tech/tags/compromise.md>), [cv](<https://devfeed.tech/tags/cv.md>), [cve](<https://devfeed.tech/tags/cve.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [github](<https://devfeed.tech/tags/github.md>), [hope](<https://devfeed.tech/tags/hope.md>), [museum](<https://devfeed.tech/tags/museum.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [upgrade](<https://devfeed.tech/tags/upgrade.md>), [utc](<https://devfeed.tech/tags/utc.md>), [vibecoding](<https://devfeed.tech/tags/vibecoding.md>), [week-notes](<https://devfeed.tech/tags/week-notes.md>), [work](<https://devfeed.tech/tags/work.md>)

### AI overview

The author's week combined security and maintenance work with renewed site contributions and personal updates. They followed the eMudhra certificate compromise and a PHP CVE, tracked Chainguard's early embargo break, and repeatedly ran `sudo apt update && sudo apt upgrade` on a Matomo deployment. They also worked on their site and Micropub endpoint, plan to be more explicit about AI use, updated their CV, and are drafting posts about backlogs and AI productivity. The post also recounts time with family, storm debris, and current reading, games, and shows.

### Source excerpt

A busy work week with: The eMudhra certificate compromise The PHP CVE Was wild keeping on top of the situtation with Chainguard breaking the embargo early, but I'd prefer to avoid saying anything too much about it publicly Spent a lot of Wednesday afternoon re-running sudo apt update && sudo apt upgrade repeatedly on my Matomo deployment And helping a friend with a site of theirs Morph's been truffling on my big fleece - thinking it's a nice blanket - which has been very cute Seems to have been one of those weeks where I'm dreaming a lot - some mostly positive dreams Was nice to get another hour's sleep last night, and being back to easier timings with us being UTC again Been nice to do a few bits on my site (and my Micropub endpoint) now I'm a bit more able to contribute to it again Being a bit more explicit with my AI usage is something I've been meaning to do - as I mostly have the metadata already - but given the reactions on Lobsters to the vibecoding tag (and some folks vocally leaving) I thought I'd be super explicit about it Did one of my regularly scheduled "update my CV" spurts - it's something I regularly recommend y'all should do Part way through a couple of blog posts, which I hope to finish in the next couple of days: Three backlogs in a trenchcoat Am I more productive using AI? Had a nice weekend with Sarah and the niblings In particular, we had a nice mosey around Wollaton Hall yesterday (the museum is so interesting) and a 😋 roast at The Railway today Stormy this weekend, so lots of (fairly minimal) debris in roads and police cars racing around Completed the GitHub Secure Fund (🤞🏼) this week Playing: (No Blue Prince) Kingshot (per usual) Apex Legends Reading: The Sins of Our Fathers Watched: The Expanse (Season 5) (Not yet: Taskmaster (Series 22)) (No Furiosa: A Mad Max Saga (2024)) The Celebrity Traitors (Series 2) The Celebrity Traitors: Uncloaked (Series 2) Ted Lasso (Season 4) Saturday Night Live (on YouTube) Saturday Night Live UK (on YouTube)

## SleepMaMi: A Universal Sleep Foundation Model for Integrating Macro- and Micro-structures

DevFeed: [SleepMaMi: A Universal Sleep Foundation Model for Integrating Macro- and Micro-structures](<https://devfeed.tech/articles/sleepmami-a-universal-sleep-foundation-model-for-integrating-macro-and-micro-structures-96815.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2602.07628>)

Author: Keondo Park, Younghoon Na, Yourim Choi, Hyunwoo Ryu, Hyun-Woo Shin, Hyung-Sin Kim

Published: 2026-10-23T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [Fine-tuning](<https://devfeed.tech/topics/fine-tuning.md>)

Tags: [analysis](<https://devfeed.tech/tags/analysis.md>), [architectures](<https://devfeed.tech/tags/architectures.md>), [autoencoder](<https://devfeed.tech/tags/autoencoder.md>), [context](<https://devfeed.tech/tags/context.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [deep-learning](<https://devfeed.tech/tags/deep-learning.md>), [dependencies](<https://devfeed.tech/tags/dependencies.md>), [design](<https://devfeed.tech/tags/design.md>), [features](<https://devfeed.tech/tags/features.md>), [foundation](<https://devfeed.tech/tags/foundation.md>), [foundation-model](<https://devfeed.tech/tags/foundation-model.md>), [foundation-models](<https://devfeed.tech/tags/foundation-models.md>), [framework](<https://devfeed.tech/tags/framework.md>), [global](<https://devfeed.tech/tags/global.md>), [hybrid](<https://devfeed.tech/tags/hybrid.md>), [learning](<https://devfeed.tech/tags/learning.md>), [medicine](<https://devfeed.tech/tags/medicine.md>), [metadata](<https://devfeed.tech/tags/metadata.md>), [model](<https://devfeed.tech/tags/model.md>), [models](<https://devfeed.tech/tags/models.md>), [sleep](<https://devfeed.tech/tags/sleep.md>)

### AI overview

SleepMaMi combines full-night sleep patterns with fine-grained biosignal features using two encoders: a Macro-Encoder models overnight temporal dependencies, while a Micro-Encoder captures short-term signal characteristics. The Macro-Encoder uses demographic-guided contrastive learning to align sleep patterns with metadata such as age, sex, and BMI; the Micro-Encoder combines masked autoencoding with multimodal contrastive learning. Pretrained on more than 20,000 polysomnography recordings (158,000 hours), the model reportedly matches or outperforms existing foundation models across varied downstream tasks, with stronger generalizability and label-efficient adaptation. The supplied text does not specify the tasks or evaluation details.

### Source excerpt

arXiv:2602.07628v3 Announce Type: replace-cross Abstract: While the shift toward unified foundation models has revolutionized many deep learning domains, sleep medicine remains largely restricted to task-specific models that focus on localized micro-structure features. These approaches often neglect the rich, multi-modal context of Polysomnography (PSG) and fail to capture the global macro-structure of a full night's sleep. To address this, we introduce SleepMaMi , a Sleep Foundation Model engineered to master both hour-long sleep architectures and fine-grained signal morphologies. Our framework utilizes a hierarchical dual-encoder design: a Macro-Encoder to model full-night temporal dependencies and a Micro-Encoder to capture short-term characteristics from biosignals. Macro-Encoder is trained via Demographic-Guided Contrastive Learning, which aligns overnight sleep patterns with objective subject metadata, such as age, sex and BMI to refine global representations. Micro-Encoder is optimized via a hybrid Masked Autoencoder (MAE) and multi-modal contrastive objective. Pre-trained on a massive corpus of $>$20,000 PSG recordings (158K hours),SleepMaMi outperforms or matches state-of-the-art existing foundation models across a diverse suite of downstream tasks, demonstrating superior generalizability and label-efficient adaptation for clinical sleep analysis.

## BrainCodes: A Lightweight EEG Foundation Model over Discrete Neural Codes

DevFeed: [BrainCodes: A Lightweight EEG Foundation Model over Discrete Neural Codes](<https://devfeed.tech/articles/braincodes-a-lightweight-eeg-foundation-model-over-discrete-neural-codes-94909.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.29320>)

Author: Kleanthis Avramidis, Ivan Tashev, Dimitra Emmanouilidou

Published: 2026-10-22T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [AI Foundation Models](<https://devfeed.tech/topics/ai-foundation-models.md>), [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [quantization](<https://devfeed.tech/topics/quantization.md>)

Tags: [architecture](<https://devfeed.tech/tags/architecture.md>), [attention](<https://devfeed.tech/tags/attention.md>), [cognitive-load](<https://devfeed.tech/tags/cognitive-load.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-ne](<https://devfeed.tech/tags/cs-ne.md>), [eeg](<https://devfeed.tech/tags/eeg.md>), [foundation-model](<https://devfeed.tech/tags/foundation-model.md>), [lightweight](<https://devfeed.tech/tags/lightweight.md>), [self-supervised-learning](<https://devfeed.tech/tags/self-supervised-learning.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [wearable](<https://devfeed.tech/tags/wearable.md>)

### AI overview

BrainCodes is a compact EEG foundation model pretrained with self-supervised learning on more than 800 hours of recordings from over 1,000 participants. It combines a quantization-based neural tokenizer with a spatiotemporal encoder designed for arbitrary electrode configurations; its training encourages spatial connectivity features and sensitivity to narrowband oscillations. The abstract reports performance matching or exceeding the state of the art on sleep staging, attention assessment, and cognitive load estimation, with transfer to low-resource wearable devices using up to 97% fewer trainable parameters than comparable architectures. The excerpt also describes evaluation of learned representation geometry and invariances across spatial, temporal, and spectral signal properties.

### Source excerpt

arXiv:2610.29320v1 Announce Type: new Abstract: Recent electroencephalography (EEG) foundation models increasingly emphasize model scale, potentially limiting their deployment on resource-constrained devices and in everyday environments. We argue that transferable EEG features can instead be learned with a compact architecture and introduce BrainCodes, a lightweight foundation model pretrained through self-supervised learning on more than 800 hours of recordings from over 1,000 participants. Our architecture combines a quantization-based neural tokenizer with a spatiotemporal encoder that supports arbitrary electrode configurations. The pretraining objective explicitly encourages representations that capture spatial connectivity through channel positional encoding, and sensitivity to narrowband neural oscillations through an auxiliary training objective. Our model matches or surpasses the state of the art across diverse neurophysiological monitoring tasks, including sleep staging, attention assessment, and cognitive load estimation. BrainCodes transfers effectively to low-resource wearable devices while using up to 97% fewer trainable parameters than comparable architectures. Beyond predictive accuracy, our evaluation framework characterizes the geometry and invariances of learned EEG representations across measurable spatial, temporal, and spectral signal properties.

## Meet Draco, the 90mm smart telescope that wakes itself up to shoot the night while you sleep

DevFeed: [Meet Draco, the 90mm smart telescope that wakes itself up to shoot the night while you sleep](<https://devfeed.tech/articles/meet-draco-the-90mm-smart-telescope-that-wakes-itself-up-to-shoot-the-night-while-you-sleep-94044.md>)

Original publisher: [Read original article](<https://www.yankodesign.com/2026/10/21/meet-draco-the-90mm-smart-telescope-that-wakes-itself-up-to-shoot-the-night-while-you-sleep/>)

Author: Sarang Sheth

Published: 2026-10-22T01:45:39Z

Content type: article

Language: en

Sources: [Yanko Design](<https://devfeed.tech/sources/yanko-design.md>)

Topics: [Image](<https://devfeed.tech/topics/image.md>), [smartdevices](<https://devfeed.tech/topics/smartdevices.md>), [User interface design](<https://devfeed.tech/topics/ui-design.md>)

Tags: [1-3](<https://devfeed.tech/tags/1-3.md>), [2](<https://devfeed.tech/tags/2.md>), [3](<https://devfeed.tech/tags/3.md>), [about](<https://devfeed.tech/tags/about.md>), [admission](<https://devfeed.tech/tags/admission.md>), [alignment](<https://devfeed.tech/tags/alignment.md>), [all](<https://devfeed.tech/tags/all.md>), [alpha](<https://devfeed.tech/tags/alpha.md>), [also](<https://devfeed.tech/tags/also.md>), [aluminum](<https://devfeed.tech/tags/aluminum.md>), [ambition](<https://devfeed.tech/tags/ambition.md>), [app](<https://devfeed.tech/tags/app.md>), [ar](<https://devfeed.tech/tags/ar.md>), [deals](<https://devfeed.tech/tags/deals.md>), [deals-product-design-technology-dwarflab-shop-telescope](<https://devfeed.tech/tags/deals-product-design-technology-dwarflab-shop-telescope.md>), [dwarflab](<https://devfeed.tech/tags/dwarflab.md>), [north](<https://devfeed.tech/tags/north.md>), [product-design](<https://devfeed.tech/tags/product-design.md>), [shop](<https://devfeed.tech/tags/shop.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [smart](<https://devfeed.tech/tags/smart.md>), [stacking](<https://devfeed.tech/tags/stacking.md>), [star](<https://devfeed.tech/tags/star.md>), [technology](<https://devfeed.tech/tags/technology.md>), [telescope](<https://devfeed.tech/tags/telescope.md>), [up](<https://devfeed.tech/tags/up.md>)

### AI overview

DWARFLAB's Draco combines a 90mm folded optical system, rotating sensor, guide camera, and onboard computing to support single exposures up to 300 seconds on an alt-azimuth mount without routine polar alignment. Its app handles target finding, focus, calibration, tracking, and stacking; scheduled sessions can run remotely or overnight when skies are clear. The 50MP camera bins pixels for nebulae and galaxies or uses native readout for solar-system targets, while a second wide-angle camera captures Milky Way views. The $1,599 SHO Edition automates narrowband Hubble Palette imaging; the $1,399 Standard Edition includes a built-in solar ND filter. Both include an Hα + O III filter, and the SHO Edition instead supplies a magnetic external ND filter for solar use. The excerpt says galaxies and Milky Way images still benefit most from dark skies; battery life is about five hours.

### Source excerpt

Meet Draco, the 90mm smart telescope that wakes itself up to shoot the night while you sleep About 4,500 years ago, while the Egyptians were busy stacking pyramids, the North Star was a completely different star called Thuban. It sat in the...

## SleepPairNet: Learning Multimodal Pediatric Sleep Representations with Random-Pair Cross-Attention

DevFeed: [SleepPairNet: Learning Multimodal Pediatric Sleep Representations with Random-Pair Cross-Attention](<https://devfeed.tech/articles/sleeppairnet-learning-multimodal-pediatric-sleep-representations-with-random-pair-cross-attention-93650.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.26628>)

Author: Saurav Raj Pandey, Harlin Lee

Published: 2026-10-21T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Transformer](<https://devfeed.tech/topics/transformer.md>)

Tags: [abstract](<https://devfeed.tech/tags/abstract.md>), [all](<https://devfeed.tech/tags/all.md>), [analysis](<https://devfeed.tech/tags/analysis.md>), [arxiv](<https://devfeed.tech/tags/arxiv.md>), [attention](<https://devfeed.tech/tags/attention.md>), [can](<https://devfeed.tech/tags/can.md>), [chat](<https://devfeed.tech/tags/chat.md>), [cross](<https://devfeed.tech/tags/cross.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [dataset](<https://devfeed.tech/tags/dataset.md>), [eeg](<https://devfeed.tech/tags/eeg.md>), [external](<https://devfeed.tech/tags/external.md>), [flexibility](<https://devfeed.tech/tags/flexibility.md>), [framework](<https://devfeed.tech/tags/framework.md>), [important](<https://devfeed.tech/tags/important.md>), [learning](<https://devfeed.tech/tags/learning.md>), [multimodal](<https://devfeed.tech/tags/multimodal.md>), [pair](<https://devfeed.tech/tags/pair.md>), [random](<https://devfeed.tech/tags/random.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [with](<https://devfeed.tech/tags/with.md>)

### AI overview

SleepPairNet learns multimodal pediatric sleep representations by using cross-attention to model pairwise interactions between physiological signals. During self-supervised pretraining, it randomly selects modality pairs, exposing the model to varied combinations and allowing flexibility across modalities at test time. Pretrained on the NCH dataset, it outperforms the evaluated self-supervised baselines on all respiratory tasks and remains competitive on EEG-related tasks; transfer to the external CHAT dataset yields stronger results. The excerpt does not specify metrics, datasets' characteristics, or which baselines and modality combinations were evaluated.

### Source excerpt

arXiv:2610.26628v1 Announce Type: new Abstract: Analyzing physiological sleep signals is important for assessing sleep quality and detecting sleep-related disorders. Learning inter-modal relationships across physiological signals can provide richer representations for downstream sleep analysis. We introduce SleepPairNet, a multimodal self-supervised learning (SSL) framework that learns pairwise interactions between physiological signals through cross-attention. We propose a random modality-pair selection strategy that exposes the model to diverse signal pairs during pretraining, enabling flexibility across modality combinations at test time. Pretrained on the NCH dataset, SleepPairNet outperforms the evaluated SSL baselines on all respiratory tasks while remaining competitive on EEG-related tasks, and shows even stronger results when transferred to the external CHAT dataset.

## Deep learning-based multi-level airway collapse site localization on obstructive sleep apnea patients based on snoring sounds

DevFeed: [Deep learning-based multi-level airway collapse site localization on obstructive sleep apnea patients based on snoring sounds](<https://devfeed.tech/articles/deep-learning-based-multi-level-airway-collapse-site-localization-on-obstructive-sleep-apnea-patients-based-on-snoring-sounds-94063.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2408.16030>)

Author: Ying-Chieh Hsu, Stanley Yung-Chuan Liu, Chao-Jung Huang, Chi-Wei Wu, Ren-Kai Cheng, Jane Yung-Jen Hsu, Shang-Ran Huang, Yuan-Ren Cheng, Fu-Shun Hsu

Published: 2026-10-21T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Deep learning](<https://devfeed.tech/topics/deep-learning.md>), [Artificial Intelligence](<https://devfeed.tech/topics/ai.md>), [Model Development](<https://devfeed.tech/topics/model-development.md>)

Tags: [ast](<https://devfeed.tech/tags/ast.md>), [audio](<https://devfeed.tech/tags/audio.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [cs-sd](<https://devfeed.tech/tags/cs-sd.md>), [deep-learning](<https://devfeed.tech/tags/deep-learning.md>), [eess-as](<https://devfeed.tech/tags/eess-as.md>), [models](<https://devfeed.tech/tags/models.md>), [osa](<https://devfeed.tech/tags/osa.md>), [recordings](<https://devfeed.tech/tags/recordings.md>), [results](<https://devfeed.tech/tags/results.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [small](<https://devfeed.tech/tags/small.md>), [study](<https://devfeed.tech/tags/study.md>), [synthetic](<https://devfeed.tech/tags/synthetic.md>), [tracking](<https://devfeed.tech/tags/tracking.md>), [transformer](<https://devfeed.tech/tags/transformer.md>), [using](<https://devfeed.tech/tags/using.md>), [vs](<https://devfeed.tech/tags/vs.md>)

### AI overview

Using snoring recordings from 37 patients who underwent drug-induced sleep endoscopy, the study trained ResNet-50 and Audio Spectrogram Transformer (AST) models to locate airway obstructions at multiple levels. AST significantly outperformed ResNet-50 and showed its strongest results for Velum, Oropharynx, and retropalatal obstructions (F1 0.59-0.87; AUC 0.87-0.97). Both models remained limited for Tongue-base, Epiglottis, and retroglossal sites, which the authors attribute to small sample sizes despite oversampling. A full-night recording analysis suggests acoustic profiling may track overnight airway dynamics, though the excerpt does not establish clinical effectiveness.

### Source excerpt

arXiv:2408.16030v3 Announce Type: replace-cross Abstract: Objective This study evaluated deep learning models, specifically ResNet-50 and Audio Spectrogram Transformer (AST), to identify multi-level upper airway collapses in obstructive sleep apnea (OSA) patients using snoring sounds. Methods Snoring recordings from 37 subjects undergoing drug-induced sleep endoscopy (DISE) between 2020 and 2021 were used to train models for two multi-label tasks: identifying obstructions at the Velum (V), Oropharynx (O), Tongue-base (T), and Epiglottis (E) levels, and the retropalatal (RP) vs. retroglossal (RG) levels. Results The AST model significantly outperformed ResNet-50 (P <0.05 after Bonferroni correction), effectively identifying V-, O-, and RP-level obstructions (F1-score: 0.59-0.87, Matthews Correlation Coefficient: 0.50-0.80, and AUC: 0.87-0.97). However, the performance of both models at the T, E, and RG levels remained limited due to small sample sizes, even with the Synthetic Minority Over-sampling Technique. A full-night recording analysis further demonstrated the feasibility of acoustic profiling for airway dynamics. Conclusion The AST models can effectively identify airway obstruction at V, O, and RP levels via snoring sounds. The proposed models offer a promising non-invasive tool for tracking overnight airway dynamics and assisting in clinical treatment planning.

## Global Time-Aware Conditioning (GTC): A Mechanism for Time-Aware Pediatric Sleep Representation Learning

DevFeed: [Global Time-Aware Conditioning (GTC): A Mechanism for Time-Aware Pediatric Sleep Representation Learning](<https://devfeed.tech/articles/global-time-aware-conditioning-gtc-a-mechanism-for-time-aware-pediatric-sleep-representation-learning-93649.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.26618>)

Author: Saurav Raj Pandey, Harlin Lee

Published: 2026-10-21T04:00:00Z

Content type: article

Language: en

Sources: [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Representation learning](<https://devfeed.tech/topics/representation-learning.md>), [Model Development](<https://devfeed.tech/topics/model-development.md>)

Tags: [architectures](<https://devfeed.tech/tags/architectures.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [gtc](<https://devfeed.tech/tags/gtc.md>), [learning](<https://devfeed.tech/tags/learning.md>), [machine-learning](<https://devfeed.tech/tags/machine-learning.md>), [mechanism](<https://devfeed.tech/tags/mechanism.md>), [self-supervised-learning](<https://devfeed.tech/tags/self-supervised-learning.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [ssl](<https://devfeed.tech/tags/ssl.md>), [tasks](<https://devfeed.tech/tags/tasks.md>), [temporal](<https://devfeed.tech/tags/temporal.md>), [time](<https://devfeed.tech/tags/time.md>), [window](<https://devfeed.tech/tags/window.md>), [windows](<https://devfeed.tech/tags/windows.md>)

### AI overview

Global Time-Aware Conditioning (GTC) adds each input window's position within the full sleep session to its latent representation, complementing local positional encodings in self-supervised sleep learning. Integrated into three representative frameworks, it was evaluated on five clinically relevant, extremely imbalanced tasks across two independent pediatric cohorts. The abstract reports general F1 and AUPRC improvements across architectures and tasks, with gains persisting in cross-cohort transfer; it provides no task-level results or implementation details.

### Source excerpt

arXiv:2610.26618v1 Announce Type: new Abstract: Sleep physiology evolves throughout the night, with changes in sleep cycles, respiratory stability, and cardiovascular dynamics. However, self-supervised learning (SSL) methods for sleep typically process short, fixed-duration windows without explicitly modeling where each window occurs within the full sleep session. We introduce Global Time-Aware Conditioning (GTC), a mechanism that complements local positional encodings by conditioning latent representations on the global time-of-sleep of each input window. We integrate GTC into three representative SSL frameworks and evaluate them on five clinically relevant tasks with extreme class imbalance across two independent pediatric cohorts. GTC generally improves F1 and AUPRC across architectures and tasks, with gains persisting under cross-cohort transfer, highlighting the value of global temporal context for sleep representation learning.

## Goals for 2025

DevFeed: [Goals for 2025](<https://devfeed.tech/articles/goals-for-2025-52484.md>)

Original publisher: [Read original article](<http://www.jackfranklin.co.uk/blog/goals-for-2025/>)

Published: 2024-12-23T00:00:00Z

Content type: article

Language: en

Sources: [Jack Franklin](<https://devfeed.tech/sources/jack-franklin.md>)

Topics: [browser-game](<https://devfeed.tech/topics/browser-game.md>), [Sports](<https://devfeed.tech/topics/sports.md>), [To-Do](<https://devfeed.tech/topics/todo.md>)

Tags: [2025](<https://devfeed.tech/tags/2025.md>), [blog](<https://devfeed.tech/tags/blog.md>), [browser](<https://devfeed.tech/tags/browser.md>), [exercise](<https://devfeed.tech/tags/exercise.md>), [fitness](<https://devfeed.tech/tags/fitness.md>), [focus](<https://devfeed.tech/tags/focus.md>), [game](<https://devfeed.tech/tags/game.md>), [goals](<https://devfeed.tech/tags/goals.md>), [health](<https://devfeed.tech/tags/health.md>), [launch](<https://devfeed.tech/tags/launch.md>), [leo](<https://devfeed.tech/tags/leo.md>), [logistics](<https://devfeed.tech/tags/logistics.md>), [network](<https://devfeed.tech/tags/network.md>), [parenting](<https://devfeed.tech/tags/parenting.md>), [side-project](<https://devfeed.tech/tags/side-project.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [steam](<https://devfeed.tech/tags/steam.md>), [v1](<https://devfeed.tech/tags/v1.md>), [weekly](<https://devfeed.tech/tags/weekly.md>), [writing](<https://devfeed.tech/tags/writing.md>)

### AI overview

After a year centered on adjusting to new parenthood, the author sets six goals for 2025: exercise at least three times a week, launch the browser-based train logistics game OnTrack on Steam, take Leo swimming monthly, continue a weekly date night, write ten blog posts, and improve productivity at work. For OnTrack, success means defining and publishing a purchasable first version, with no sales target; the author may use Tauri to package the browser game. The goals reflect hopes that nursery will free time for exercise and work, and a desire to protect time for family, their relationship, and writing.

### Source excerpt

As 2024 comes to a close, I wanted to take a moment to reflect on what has been the most incredible and challenging year of my life. In January our child Leo was born and our lifes as we knew them were changed forever. There is plenty to be written on the highs and lows of being new parents and without doubt it has been the most difficult and rewarding year of my life. This year it has been impossible to set routines and goals because our sole focus was Leo and adapting to parenting. Any other goals or accomplishments were secondary and whilst to some extent that will always be the case, I wanted to note down some goals I have for 2025 to hold myself accountable. Hopefully this time next year I'm smiling writing a post about how successful I've been! Exercise at least 3 times a week My health, both physically and mentally, has suffered this year from a lack of time to invest in myself. Weekly badminton sessions and long walks with Leo and the dog have kept things ticking over, but the lack of sleep, quick meals once Leo is asleep and having to balance parenting and returning back to work has had a negative impact on my fitness. With Leo at nursery in 2025, I hope to have more time during the week to exercise, and I'm starting with a goal of 3 times per week. I hope to do more, but if I can achieve this it will be a solid improvement over 2024. Launch my side project on Steam I've been working on OnTrack for over two years. It's a browser based train logistics game where you have to build a passenger network and balance customer satisfaction and finances. At times I have suffered from focusing on too many features and knowing how to wrap up a "v1". This year, I want to define and launch the first version. My definition of success here is that the game is on Steam and you can buy it. I have no expectation on sales or making any money, but it would be a huge achievement to meet this milestone. Because the game is browser based (I did venture into Rust before falling ba

## Sleepless Concurrency: delay() vs. Thread.sleep()

DevFeed: [Sleepless Concurrency: delay() vs. Thread.sleep()](<https://devfeed.tech/articles/sleepless-concurrency-delay-vs-thread-sleep-25750.md>)

Original publisher: [Read original article](<https://blog.shreyaspatil.dev/sleepless-concurrency-delay-vs-threadsleep/>)

Author: Shreyas Patil

Published: 2023-07-31T04:39:30Z

Content type: article

Language: en

Sources: [Shreyas Patil's Blog](<https://devfeed.tech/sources/shreyas-patil-s-blog.md>)

Topics: [Coroutines](<https://devfeed.tech/topics/coroutines.md>), [Kotlin](<https://devfeed.tech/topics/kotlin.md>), [Distributed Systems & Parallel Computing](<https://devfeed.tech/topics/distributed-systems-parallel-computing.md>)

Tags: [android](<https://devfeed.tech/tags/android.md>), [asynchronous](<https://devfeed.tech/tags/asynchronous.md>), [blocking](<https://devfeed.tech/tags/blocking.md>), [concurrency](<https://devfeed.tech/tags/concurrency.md>), [coroutine](<https://devfeed.tech/tags/coroutine.md>), [coroutines](<https://devfeed.tech/tags/coroutines.md>), [delay](<https://devfeed.tech/tags/delay.md>), [java](<https://devfeed.tech/tags/java.md>), [jvm](<https://devfeed.tech/tags/jvm.md>), [kotlin](<https://devfeed.tech/tags/kotlin.md>), [kotlin-coroutines](<https://devfeed.tech/tags/kotlin-coroutines.md>), [main-thread](<https://devfeed.tech/tags/main-thread.md>), [multithreading](<https://devfeed.tech/tags/multithreading.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [thread](<https://devfeed.tech/tags/thread.md>)

### AI overview

On a single thread, coroutine `delay()` suspends the coroutine without blocking the thread, allowing other coroutines to run; Java `Thread.sleep()` blocks the thread, preventing them from running until sleep ends. In the excerpt's two-coroutine example with a 1,000 ms wait, both coroutines launch in the same second with `delay()`, and total execution takes 1,045 ms; with `sleep()`, the second launches about one second later and total execution takes 2,044 ms. The excerpt frames the APIs as serving different use cases.

### Source excerpt

Explore the difference between delay() and Thread.sleep() in Kotlin. Understand how non-blocking concurrency works from a platform perspective.

## Connect to Open VPN during Gitlab Pipeline

DevFeed: [Connect to Open VPN during Gitlab Pipeline](<https://devfeed.tech/articles/connect-to-open-vpn-during-gitlab-pipeline-25555.md>)

Original publisher: [Read original article](<https://www.marcogomiero.com/posts/2021/gitlab-pipeline-vpn/>)

Author: Marco Gomiero

Published: 2021-06-28T00:00:00Z

Content type: tutorial

Language: en

Sources: [Posts on Marco Gomiero](<https://devfeed.tech/sources/posts-on-marco-gomiero.md>)

Topics: [gitlab-ci](<https://devfeed.tech/topics/gitlab-ci.md>), [CI/CD](<https://devfeed.tech/topics/cicd.md>), [Virtual Private Network](<https://devfeed.tech/topics/vpn.md>)

Tags: [ci](<https://devfeed.tech/tags/ci.md>), [ci-cd](<https://devfeed.tech/tags/ci-cd.md>), [dependencies](<https://devfeed.tech/tags/dependencies.md>), [download](<https://devfeed.tech/tags/download.md>), [fail](<https://devfeed.tech/tags/fail.md>), [gitlab](<https://devfeed.tech/tags/gitlab.md>), [kotlin](<https://devfeed.tech/tags/kotlin.md>), [libraries](<https://devfeed.tech/tags/libraries.md>), [maven](<https://devfeed.tech/tags/maven.md>), [maven-repository](<https://devfeed.tech/tags/maven-repository.md>), [openvpn](<https://devfeed.tech/tags/openvpn.md>), [password](<https://devfeed.tech/tags/password.md>), [pipeline](<https://devfeed.tech/tags/pipeline.md>), [script](<https://devfeed.tech/tags/script.md>), [secrets](<https://devfeed.tech/tags/secrets.md>), [server](<https://devfeed.tech/tags/server.md>), [sleep](<https://devfeed.tech/tags/sleep.md>), [type](<https://devfeed.tech/tags/type.md>), [vpn](<https://devfeed.tech/tags/vpn.md>)

### AI overview

GitLab CI pipelines can connect to a VPN before running stages so they can download dependencies from a private Maven repository behind it. The excerpt uses OpenVPN and requires three CI/CD secret variables: CLIENT_OVPN for the .ovpn file contents, VPN_USER, and VPN_PWD. It describes adding setup commands before the stages, then waiting 30 seconds and pinging a server to check the connection. The author says the script can be adapted for other VPN types, but the excerpt omits the actual commands and complete configuration.

### Source excerpt

Gitlab CI/CD offers the possibility to create a pipeline, which runs when something changes in the repository. A pipeline consists of one or more stages that run in order and in these stages, for example, it is possible to build the project, run the tests, create the artifacts, etc. For more information about Gitlab CI/CD, I suggest you look over the documentation. These out-of-the-box solutions really simplify the work to be done to have a CI up and running. For example, this is the configuration file (a file called .gitlab-ci.yml placed at the repository's root) needed for running all the tests of a Kotlin project.