Open-set semi-supervised learning (OSSL) leverages unlabeled data containing both in-distribution (ID) and unknown out-of-distribution (OOD) samples, aiming simultaneously to improve closed-set accuracy and detect novel OOD instances. Existing methods either discard valuable information from uncertain samples or force-align every unlabeled sample into one or a few synthetic “catch-all” representations, resulting in geometric collapse and overconfidence on only seen OODs. To address the limitations, we introduce selective non-alignment, adding a novel “skip” operator into conventional pull and push operations of contrastive learning. Our framework, SkipAlign, selectively skips alignment (pulling) for low-confidence unlabeled samples, retaining only gentle repulsion against ID prototypes. This approach transforms uncertain samples into a pure repulsion signal, resulting in tighter ID clusters and naturally dispersed OOD features. Extensive experiments demonstrate that SkipAlign significantly outperforms state-of-the-art methods in detecting unseen OOD data without sacrificing ID classification accuracy.

As AI moves from data centers to robots and wearables, scaling ever-larger models becomes insufficient. Physical AI operates under tight latency, energy, privacy, and reliability constraints, and its performance depends not only on model capacity but also on how signals are acquired through controllable sensors in dynamic environments. We present Artificial Tripartite Intelligence (ATI), a bio-inspired, sensor-first architectural contract for physical AI. ATI is tripartite at the systems level: a Brainstem (L1) provides reflexive safety and signal-integrity control, a Cerebellum (L2) performs continuous sensor calibration, and a Cerebral Inference Subsystem spanning L3/L4 supports routine skill selection and execution, coordination, and deep reasoning. This modular organization allows sensor control, adaptive sensing, edge-cloud execution, and foundation model reasoning to co-evolve within one closed-loop architecture, while keeping time-critical sensing and control on device and invoking higher-level inference only when needed. We instantiate ATI in a mobile camera prototype under dynamic lighting and motion. In our routed evaluation (L3-L4 split inference), compared to the default auto-exposure setting, ATI (L1/L2 adaptive sensing) improves end-to-end accuracy from 53.8% to 88% while reducing remote L4 invocations by 43.3%. These results show the value of co-designing sensing and inference for embodied AI.

Sleep is essential for health and well-being, yet the gold-standard method for sleep stage assessment, polysomnography (PSG), requires overnight monitoring with numerous wired sensors and manual scoring, limiting its scalability and comfort in real-world settings. We present ViSS, the first end-to-end deep learning framework for fully non-contact sleep staging directly from raw infrared video. To capture the dual temporal structure of sleep, ViSS integrates per-epoch motion encoding and long-range sequence modeling in a compact hierarchical architecture, without relying on intermediate physiological signals. Evaluated on the largest infrared video dataset to date, including both healthy individuals and patients, ViSS achieves 80% accuracy and 0.78 macro F1-score using fewer than 7 million parameters. These results establish infrared video as a viable modality for automated sleep staging and highlight the potential of ViSS for scalable, privacy-preserving home-based assessment.

Obstructive sleep apnea (OSA) is a prevalent sleep disorder, affecting approximately one billion individuals globally. In this study, we aim to address the limitations of Polysomnography (PSG), the gold standard for OSA diagnosis, by developing SlAction, a non-intrusive system that utilizes infrared videos for OSA detection in daily sleep settings. Considering the privacy-sensitive nature of sleep videos, SlAction is designed to analyze data directly on the camera-capturing device, eliminating the need to transmit video data to a server. With the collaboration of clinical experts, we extensively analyze the largest dataset worldwide that we collected, establishing correlations between OSA events and human motions during sleep. Our novel approach achieved an OSA prediction performance with an F1 score of 0.88. Notably, even when running on a low-spec CPU, our SlAction operates approximately 75 times faster than previous work evaluated on high-performance GPU servers.

While federated learning leverages distributed client resources, it faces challenges due to heterogeneous client capabilities. This necessitates allocating models suited to clients' resources and careful parameter aggregation to accommodate this heterogeneity. We propose HypeMeFed, a novel federated learning framework for supporting client heterogeneity by combining a multi-exit network architecture with hypernetwork-based model weight generation. This approach aligns the feature spaces of heterogeneous model layers and resolves per-layer information disparity during weight aggregation. To practically realize HypeMeFed, we also propose a low-rank factorization approach to minimize computation and memory overhead associated with hypernetworks. Our evaluations on a real-world heterogeneous device testbed indicate that HypeMeFed enhances accuracy by 5.12% over FedAvg, reduces the hypernetwork memory requirements by 98.22%, and accelerates its operations by 1.86X compared to a naive hypernetwork approach. These results demonstrate HypeMeFed's effectiveness in leveraging and engaging heterogeneous clients for federated learning.

Running deep learning models on resource-constrained edge devices has drawn significant attention due to its fast response, privacy preservation, and robust operation regardless of Internet connectivity. While these devices already cope with various intelligent tasks, the latest edge devices that are equipped with multiple types of low-power accelerators (i.e., both mobile GPU and NPU) can bring another opportunity; a task that used to be too heavy for an edge device in the single-accelerator world might become viable in the upcoming heterogeneous-accelerator world. To realize the potential in the context of 3D object detection, we identify several technical challenges and propose PointSplit, a novel 3D object detection framework for multi-accelerator edge devices that addresses the problems. Specifically, our PointSplit design includes (1) 2D semantics-aware biased point sampling, (2) parallelized 3D feature extraction, and (3) role-based group-wise quantization. We implement PointSplit on TensorFlow Lite and evaluate it on a customized hardware platform comprising both mobile GPU and EdgeTPU. Experimental results on representative RGB-D datasets, SUN RGB-D and Scannet V2, demonstrate that PointSplit on a multi-accelerator device is 24.7 × faster with similar accuracy compared to the full-precision, 2D-3D fusion-based 3D detector on a GPU-only device.