Abstract
Speech self-supervised learning aims to learn general-purpose representations for downstream speech tasks. However, current approaches rely on complex, carefully designed prediction targets. We challenge this necessity with GLaS-JEPA, a framework that directly predicts the current encoder's continuous representations at masked positions, without contrastive learning, discrete targets, or separate EMA target encoders. We prevent representation collapse using SIGReg representation-space regularization, eliminating the need for engineered target-generation mechanisms. Pretrained on 960 hours of LibriSpeech, our 57M-parameter model achieves a 6.89% WER on frozen-encoder SUPERB ASR and a 25.87% CER on slot filling, outperforming the best non-distilled sub-90M baselines by 43.1% and 22.0%, respectively. These results demonstrate that highly competitive speech representations can emerge from a radically simplified training recipe.
Keywords
Subject
Publication details
- Journal
- Not available
- Open access
- Green open access
Cite this article
APA 7
Botté, G., Baroudi, S., Sadok, S., Paissan, F., Hueber, T., Alameda-Pineda, X., Marxer, R., & Ravanelli, M. (2026). GLaS-JEPA: Gaussian-Regularized Speech SSL without Engineered Prediction Targets. https://omanscience.com/en/articles/glas-jepa-gaussian-regularized-speech-ssl-without-engineered-prediction-targets
MLA 9
Botté, Gaspard, et al. "GLaS-JEPA: Gaussian-Regularized Speech SSL without Engineered Prediction Targets." https://omanscience.com/en/articles/glas-jepa-gaussian-regularized-speech-ssl-without-engineered-prediction-targets.
Chicago (author–date)
Botté, Gaspard, Séverin Baroudi, Samir Sadok, Francesco Paissan, Thomas Hueber, Xavier Alameda-Pineda, Ricard Marxer, and Mirco Ravanelli. 2026. "GLaS-JEPA: Gaussian-Regularized Speech SSL without Engineered Prediction Targets." https://omanscience.com/en/articles/glas-jepa-gaussian-regularized-speech-ssl-without-engineered-prediction-targets.
Harvard
Botté, G., Baroudi, S., Sadok, S., Paissan, F., Hueber, T., Alameda-Pineda, X., Marxer, R. and Ravanelli, M. (2026) 'GLaS-JEPA: Gaussian-Regularized Speech SSL without Engineered Prediction Targets', Available at: https://omanscience.com/en/articles/glas-jepa-gaussian-regularized-speech-ssl-without-engineered-prediction-targets.
Vancouver
Botté G, Baroudi S, Sadok S, Paissan F, Hueber T, Alameda-Pineda X, et al. GLaS-JEPA: Gaussian-Regularized Speech SSL without Engineered Prediction Targets. https://omanscience.com/en/articles/glas-jepa-gaussian-regularized-speech-ssl-without-engineered-prediction-targets
IEEE
G. Botté, S. Baroudi, S. Sadok, F. Paissan, T. Hueber, X. Alameda-Pineda, R. Marxer, and M. Ravanelli, "GLaS-JEPA: Gaussian-Regularized Speech SSL without Engineered Prediction Targets," https://omanscience.com/en/articles/glas-jepa-gaussian-regularized-speech-ssl-without-engineered-prediction-targets.