Soft Clustering Anchors for Self-Supervised Speech Representation Learning in Joint Embedding Prediction Architectures

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Ioannides, Georgios, Kieback, Adrian, Goldfeder, Judah, Pang, Linsey, Chadha, Aman, Elkins, Aaron, LeCun, Yann, Shwartz-Ziv, Ravid
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908823337304064
author Ioannides, Georgios
Kieback, Adrian
Goldfeder, Judah
Pang, Linsey
Chadha, Aman
Elkins, Aaron
LeCun, Yann
Shwartz-Ziv, Ravid
author_facet Ioannides, Georgios
Kieback, Adrian
Goldfeder, Judah
Pang, Linsey
Chadha, Aman
Elkins, Aaron
LeCun, Yann
Shwartz-Ziv, Ravid
contents Joint Embedding Predictive Architectures (JEPA) offer a promising approach to self-supervised speech representation learning, but suffer from representation collapse without explicit grounding. We propose GMM-Anchored JEPA, which fits a Gaussian Mixture Model once on log-mel spectrograms and uses its frozen soft posteriors as auxiliary targets throughout training. A decaying supervision schedule allows GMM regularization to dominate early training before gradually yielding to the JEPA objective. Unlike HuBERT and WavLM, which require iterative re-clustering, our approach clusters input features once with soft rather than hard assignments. On ~50k hours of speech, GMM anchoring improves ASR (28.68% vs. 33.22% WER), emotion recognition (67.76% vs. 65.46%), and slot filling (64.7% vs. 59.1% F1) compared to a WavLM-style baseline with matched compute. Cluster analysis shows GMM-anchored representations achieve up to 98% entropy compared to 31% for WavLM-style, indicating substantially more uniform cluster utilization. Code is made available at https://github.com/gioannides/clustering-anchored-jepa.
format Preprint
id arxiv_https___arxiv_org_abs_2602_09040
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Soft Clustering Anchors for Self-Supervised Speech Representation Learning in Joint Embedding Prediction Architectures
Ioannides, Georgios
Kieback, Adrian
Goldfeder, Judah
Pang, Linsey
Chadha, Aman
Elkins, Aaron
LeCun, Yann
Shwartz-Ziv, Ravid
Audio and Speech Processing
Artificial Intelligence
Machine Learning
Sound
Joint Embedding Predictive Architectures (JEPA) offer a promising approach to self-supervised speech representation learning, but suffer from representation collapse without explicit grounding. We propose GMM-Anchored JEPA, which fits a Gaussian Mixture Model once on log-mel spectrograms and uses its frozen soft posteriors as auxiliary targets throughout training. A decaying supervision schedule allows GMM regularization to dominate early training before gradually yielding to the JEPA objective. Unlike HuBERT and WavLM, which require iterative re-clustering, our approach clusters input features once with soft rather than hard assignments. On ~50k hours of speech, GMM anchoring improves ASR (28.68% vs. 33.22% WER), emotion recognition (67.76% vs. 65.46%), and slot filling (64.7% vs. 59.1% F1) compared to a WavLM-style baseline with matched compute. Cluster analysis shows GMM-anchored representations achieve up to 98% entropy compared to 31% for WavLM-style, indicating substantially more uniform cluster utilization. Code is made available at https://github.com/gioannides/clustering-anchored-jepa.
title Soft Clustering Anchors for Self-Supervised Speech Representation Learning in Joint Embedding Prediction Architectures
topic Audio and Speech Processing
Artificial Intelligence
Machine Learning
Sound
url https://arxiv.org/abs/2602.09040