# MCR-Data2vec 2.0: Improving Self-supervised Speech Pre-training via Model-level Consistency Regularization

arXiv: https://arxiv.org/abs/2306.08463 · HTML: https://mlsanity.com/paper/2306.08463/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 14 Jun 2023.

## Built on by (most cited first, 3 of 3)

- [PFML: Self-Supervised Learning of Time-Series Data Without Representation Collapse](https://mlsanity.com/paper/2411.10087/) (2024)
- [Benchmarking Prosody Encoding in Discrete Speech Tokens](https://mlsanity.com/paper/2508.11224/) (2025)
- [GLaS-JEPA: Gaussian-Regularized Speech SSL without Engineered Prediction Targets](https://mlsanity.com/paper/2609.37798/) (2026)

## Abstract (arXiv)

Self-supervised learning (SSL) has shown significant progress in speech processing tasks. However, despite the intrinsic randomness in the Transformer structure, such as dropout variants and layer-drop, improving the model-level consistency remains under-explored in the speech SSL literature. To address this, we propose a new pre-training method that uses consistency regularization to improve Data2vec 2.0, the recent state-of-the-art (SOTA) SSL model. Specifically, the proposed method involves sampling two different student sub-models within the Data2vec 2.0 framework, enabling two output variants derived from a single input without additional parameters. Subsequently, we regularize the outputs from the student sub-models to be consistent and require them to predict the representation of the teacher model. Our experimental results demonstrate that the proposed approach improves the SSL model's robustness and generalization ability, resulting in SOTA results on the SUPERB benchmark.