🚗✨ How nice would it be to compress robot intelligence without compressing its performance?

📄 Explore the method:

👉 t.co/kB9GcyBUOT

#Robotics #EmbeddedAI #MachineLearning #OODDetection #VAE #AutonomousSystems #JetsonNano t.co/g5TUt5BTRl

Researchers introduced EXPLOR, expanding latent space with pseudo‑labeling to improve OOD uncertainty rejection, achieving accuracy on benchmarks. Read more: https://getnews.me/explor-framework-improves-out-of-distribution-uncertainty-rejection/ #explor #ooddetection #machinelearning

EMOE (Expansive Matching of Experts) improves OOD rejection with a matching loss, achieving higher accuracy on image, tabular and text benchmarks. Read more: https://getnews.me/emoe-framework-boosts-out-of-distribution-uncertainty-rejection/ #ooddetection #machinelearning

ANTS, a training‑free, zero‑shot method using multimodal LLMs, cut the false‑positive rate by 4.2% at 95% recall on ImageNet. The paper was released September 11 2025. Read more: https://getnews.me/ants-method-uses-multimodal-llms-to-boost-ood-detection/ #ants #ooddetection #multimodalllm

SCOPED detects OOD in diffusion models with a single forward pass and a Jacobian‑vector product, matching top precision‑recall while using about one‑tenth the compute. Read more: https://getnews.me/scoped-fast-out-of-distribution-detection-for-diffusion-models-2/ #scoped #ooddetection

SCOPED flags OOD inputs with a single forward pass and one Jacobian‑vector product, cutting passes by about ten‑fold while matching benchmark precision‑recall on four vision datasets. Read more: https://getnews.me/scoped-fast-out-of-distribution-detection-for-diffusion-models/ #scoped #ooddetection

Task‑aware OOD detection uses a joint likelihood of input and model prediction via a score‑based diffusion model, correlating with error on PDE, satellite and brain‑tumor data. Read more: https://getnews.me/task-aware-ood-detection-advances-trust-in-scientific-ai/ #ooddetection #scientificai

NERO (Neuron-level Relevance for Out-of-Distribution) improves OOD detection in medical imaging, beating state-of-the-art on the Kvasir and GastroVision GI benchmarks (24 Sept 2025). Read more: https://getnews.me/nero-improves-out-of-distribution-detection-in-medical-imaging/ #nero #ooddetection

The study evaluated eight methods on ultrasound tasks; image tasks improved OOD detection for lighting shifts, organ‑level tasks helped anatomical shifts. Read more: https://getnews.me/classification-tasks-influence-ood-detection-in-fetal-ultrasound-ai/ #ooddetection #fetalultrasound

The framework merges prompt optimization with subspace learning to boost few‑shot OOD detection in vision‑language models, beating prior detectors on real‑world datasets. Read more: https://getnews.me/prompt-optimization-meets-subspace-learning-for-few-shot-ood-detection/ #ooddetection #vlm

Controlling Neural Collapse boosts OOD detection and transfer‑learning, the study shows. Presented at ICML 2025, it was tested on CIFAR‑10‑C, SVHN and ImageNet‑O. Read more: https://getnews.me/controlling-neural-collapse-boosts-ood-detection-and-transfer-learning/ #neuralcollapse #ooddetection

RSCL adds class‑wise temperature scaling and mines informative outliers, improving OOD detection and tail‑class accuracy versus standard SCL on benchmarks. Read more: https://getnews.me/refined-class-learning-improves-ood-detection-on-long-tailed-data/ #ooddetection #longtailed