AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

SPEAR: A Unified SSL Framework for Learning Speech and Audio Representations

Xiaoyu Yang, Yifan Yang, Zengrui Jin, Ziyun Cui, Wen Wu, Baoxiang Li, Chao Zhang, Phil Woodland

ICML 2026 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Self-supervised speech representations

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add token mixing augmentation (ablation)SPEAR-Speech Base · SUPERB SDDER: 4.23 -> 3.01Improved
add token mixing augmentation (ablation)SPEAR-Speech Base · SUPERB SSSI-SDRi: 10.52 -> 11.17Improved
add token mixing augmentation (ablation)SPEAR-Speech Base · LibriSpeech LS-100WER: 3.1 -> 3.1 (clean)No clear effect
add token mixing augmentation (ablation)SPEAR-Speech Base · SUPERB PRPER: 3.3 -> 3.4No clear effect
use Asymmetrical instead of Disjoint dual-domain loss (ablation)SPEAR-SA Large · HEARaverage score: 78.3 -> 79.0Improved
use Asymmetrical instead of Joint dual-domain loss (ablation)SPEAR-SA Large · HEARaverage score: 78.7 -> 79.0Improved
pre-train on speech and audio jointly instead of speech onlySPEAR Large · HEARaverage score: 72.96 -> 79.26Improved
pre-train on speech and audio jointly instead of speech onlySPEAR Large · LibriSpeech LS-960 test-otherWER: 3.3 -> 3.4Hurt
use MVQ tokens instead of k-means targets (ablation)SPEAR · SUPERB SID, ERaccuracyImproved
replace Zipformer encoder with standard Transformer (ablation)SPEAR · ASR fine-tuningefficiency and WERHurt
increase number of MVQ codebooks (ablation)SPEAR · speech vs general audiodownstream performanceMixed
scale to XLarge with 197k hoursSPEAR-SA · AudioSet AS-2MmAP: 49.6 -> 50.0Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy