Whole-Study MIL Improves Prenatal CHD Screening
Masked-autoencoder pre-training, disease-robust cardiac-frame gating, and transformer pooling produce study-level CHD predictions without clinician-selected frames.
Oct 10, 20264 min2609.31376
Electrical Engineering
Signal processing, systems and control, audio and image processing.
Sort:
HuPER Uses Self-Learning to Improve Low-Resource Phonetic Recognition
A human-inspired framework combines acoustic evidence with linguistic knowledge, reporting state-of-the-art phonetic error rates from 100 hours of training data and zero-shot results in 95 unseen languages.
Oct 6, 20262 min2602.01634 Code available
Pruned CTC Cuts 180K-Vocabulary ASR Training Memory
Restricting CTC alignment states to batch target tokens preserves full-vocabulary loss and gradients while cutting full-step memory 5.1× at 180K tokens.
Oct 4, 20264 min2609.33645
Dual Transcripts Expand Minnan Speech Training Data
WenetSpeech-Min pairs Minnan and Mandarin text for around 10,000 hours of online speech, creating shared training and evaluation material for ASR and TTS.
Oct 1, 20264 min2609.36834
Symbolic Planning Lifts YuE2’s Full-Song Musicality
A shared autoregressive–non-autoregressive model writes lead sheets before audio tokens, raising expert overall-quality preferences to 49.3% versus 34.6% without planning.
Sep 29, 20264 min2609.33757
Consensus Alignment Improves Child-Speech ASR Training Data
BEACON ensembles word timestamps from off-the-shelf ASR systems to curate 413 hours of CHILDES audio, reducing average out-of-domain WER by up to 19.5%.
Sep 26, 20264 min2607.03670
ORION-CMR Delivers On-Scanner Cardiac MRI Reports in 90 Seconds
A pretrained CMR foundation model combines image analysis with local language-model reporting, reaching 81.4% expert agreement across a 68-subject clinical cohort.
Sep 24, 20264 min2609.23950
DICOM Packaging Extends NLST CT Annotations to 26,194 Patients
Three IDC analysis-result datasets combine standardized DICOM segmentations, measurements, and viewers, making NLST-derived annotations queryable without local image-processing infrastructure.
Sep 21, 20264 min2609.10858 Code available
Explicit Planning Improves Long-Form Music Generation Scores
An ABC notation plan edited by a text LLM raises SongBench’s mean score from 6.6438 to 6.7220 before token-level audio rendering.
Sep 19, 20264 min2609.16034
Bidirectional Memory Improves Thoracic Contouring and Clinical Workflow
An anatomy-change-aware 2.5D model uses bidirectional slice memory and uncertainty-guided review to cut reader contouring time by 75–80% in a 17-oncologist study.
Sep 16, 20264 min2609.16036