SpidR: Learning Fast and Stable Linguistic Units for Spoken Language Models Without Supervision

Maxime Poli, Mahi Luthra, Youssef Benchekroun, Yosuke Higuchi, Martin Gleize, Jiayi Shen, Robin Algayres, Yu-An Chung, Mido Assran, Juan Pino, Emmanuel Dupoux

Abstract: The parallel advances in language modeling and speech representation learning have raised the prospect of learning language directly from speech without textual intermediates. This requires extracting semantic representations directly from speech. Our contributions are threefold. First, we introduce SpidR, a self-supervised speech representation model that efficiently learns representations with highly accessible phonetic information, which makes it particularly suited for textless spoken language modeling. It is trained on raw waveforms using a masked prediction objective combined with self-distillation and online clustering. The intermediate layers of the student model learn to predict assignments derived from the teacher's intermediate layers. This learning objective stabilizes the online clustering procedure compared to previous approaches, resulting in higher quality codebooks. SpidR outperforms wav2vec 2.0, HuBERT, WavLM, and DinoSR on downstream language modeling benchmarks (sWUGGY, sBLIMP, tSC). Second, we systematically evaluate across models and layers the correlation between speech unit quality (ABX, PNMI) and language modeling performance, validating these metrics as reliable proxies. Finally, SpidR significantly reduces pretraining time compared to HuBERT, requiring only one day of pretraining on 16 GPUs, instead of a week. This speedup is enabled by the pretraining method and an efficient codebase, which allows faster iteration and easier experimentation. We open-source the training code and model checkpoints at https://github.com/facebookresearch/spidr.

Submission Length: Regular submission (no more than 12 pages of main content)

Changes Since Last Submission:

- Revised the abstract
- More context on SSL, ASR and downstream tasks, unsupersived pattern discovery in Related Work
- Clarified that we are not doing speech (re)synthesis
- Added discussion on consequences of our findings on models based on LLM + adapter in Conclusion.
- Added a table summarizing the metrics used in the paper.

- Fixed typos in the abstract

Code: https://github.com/facebookresearch/spidr

Assigned Action Editor: ~Tatiana_Likhomanenko1

Submission Number: 5256