mlsanityData to 14 Oct 2026

GLaS-JEPA: Gaussian-Regularized Speech SSL without Engineered Prediction Targets

arXiv 2609.37798 ↗First posted 29 Sep 2026eess.AS cs.AI cs.LGcited by 0

Not on the map. Why?

Abstract (arXiv)

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.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 27 of its 29 references that are in the map. First posted on arXiv 29 Sep 2026.

Builds on

27 of 29 references.

References · 27

  1. 2020wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representationscited by 3,905
  2. 2020Exploring Simple Siamese Representation Learningcited by 2,866
  3. 2021HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Unitscited by 2,342
  4. 2020Conformer: Convolution-augmented Transformer for Speech Recognitioncited by 1,863
  5. 2021Barlow Twins: Self-Supervised Learning via Redundancy Reductioncited by 1,838
  6. 2021WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processingcited by 1,657
  7. 2021VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learningcited by 1,057
  8. 2023Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecturecited by 1,039
  9. 2021SUPERB: Speech processing Universal PERformance Benchmarkcited by 666
  10. 2019vq-wav2vec: Self-Supervised Learning of Discrete Speech Representationscited by 340
  11. 2026LeWorldModel: Stable End-to-End Joint-Embedding Predictive Architecture from Pixelscited by 273
  12. 2025LeJEPA: Provable and Scalable Self-Supervised Learning Without the Heuristicscited by 254
15 more
  1. 2020TERA: Self-Supervised Learning of Transformer Encoder Representation for Speechcited by 188
  2. 2019An Unsupervised Autoregressive Model for Speech Representation Learningcited by 186
  3. 2022Self-supervised Learning with Random-projection Quantizer for Speech Recognitioncited by 162
  4. 2022RankMe: Assessing the downstream performance of pretrained self-supervised representations by their rankcited by 153
  5. 2020Multi-task self-supervised learning for Robust Speech Recognitioncited by 130
  6. 2022Efficient Self-supervised Learning with Contextualized Target Representations for Vision, Speech and Languagecited by 110
  7. 2020Unsupervised pretraining transfers well across languagescited by 101
  8. 2020DeCoAR 2.0: Deep Contextualized Acoustic Representations with Vector Quantizationcited by 77
  9. 2023A-JEPA: Joint-Embedding Predictive Architecture Can Listencited by 53
  10. 2026When Does LeJEPA Learn a World Model?cited by 30
  11. 2025Audio-JEPA: Joint-Embedding Predictive Architecture for Audio Representation Learningcited by 17
  12. 2025USAD: Universal Speech and Audio Representation via Distillationcited by 7
  13. 2025WavJEPA: Semantic learning unlocks robust audio foundation models for raw waveformscited by 7
  14. 2023MCR-Data2vec 2.0: Improving Self-supervised Speech Pre-training via Model-level Consistency Regularizationcited by 3
  15. 2026S-JEPA : Soft Clustering Anchors for Self-Supervised Speech Representation Learningcited by 2

Built on by

Not cited yet.