Gumbel-BEARD: Automated Layer Selection Framework Improves Whisper Speech Recognition in Low-Resource Domains
Researchers have developed Gumbel-BEARD, a domain adaptation framework that automatically selects optimal encoder layers in the Whisper speech model to improve performance in low-resource settings. The method uses a self-supervised approach with a BEST-RQ objective that adapts to target acoustic characteristics without manual tuning. The approach achieves state-of-the-art results on multiple datasets, matching fully supervised baselines with significantly less labeled data.
Gumbel-BEARD addresses a key challenge in speech recognition: adapting foundation models like Whisper to specialized domains where training data is scarce. The framework uses an end-to-end trainable hard Gumbel-Softmax selector to automatically choose which encoder layers to use for adaptation, eliminating the need for manual hyperparameter tuning. Testing on the MyST child speech corpus showed the method could match performance of models trained on 133 hours of labeled data using only 10 hours of labeled data for fine-tuning. The researchers achieved new state-of-the-art word error rates of 8.21% on MyST with Whisper-medium and 11.06% on the OGI Spontaneous dataset with Whisper-small. Additional evaluation on the CORAAL dataset demonstrated robustness to dialectal variations, with up to 6% relative word error rate reduction, suggesting the approach generalizes well across diverse low-resource conditions.
What different sources said
- arXiv cs.CLCenter
Gumbel-BEARD: Automatic Layer Selection for Self-Supervised Adaptation of Whisper in Low-Resource Domains
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.