pith. sign in

arxiv: 2111.09296 · v3 · pith:DYZIS3MHnew · submitted 2021-11-17 · 💻 cs.CL · cs.SD· eess.AS

XLS-R: Self-supervised Cross-lingual Speech Representation Learning at Scale

classification 💻 cs.CL cs.SDeess.AS
keywords speechxls-rlanguagescross-lingualpretrainingaveragedataenglish
0
0 comments X
read the original abstract

This paper presents XLS-R, a large-scale model for cross-lingual speech representation learning based on wav2vec 2.0. We train models with up to 2B parameters on nearly half a million hours of publicly available speech audio in 128 languages, an order of magnitude more public data than the largest known prior work. Our evaluation covers a wide range of tasks, domains, data regimes and languages, both high and low-resource. On the CoVoST-2 speech translation benchmark, we improve the previous state of the art by an average of 7.4 BLEU over 21 translation directions into English. For speech recognition, XLS-R improves over the best known prior work on BABEL, MLS, CommonVoice as well as VoxPopuli, lowering error rates by 14-34% relative on average. XLS-R also sets a new state of the art on VoxLingua107 language identification. Moreover, we show that with sufficient model size, cross-lingual pretraining can outperform English-only pretraining when translating English speech into other languages, a setting which favors monolingual pretraining. We hope XLS-R can help to improve speech processing tasks for many more languages of the world.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 21 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. MixFake: Benchmarking and Enhancing Audio Deepfake Detection in Diverse Real-world Mixed Audio

    cs.SD 2026-05 unverdicted novelty 7.0

    MixFake is a new benchmark for mixed-authenticity audio and a multi-stream prompt tuning method achieves 0.95% EER foreground and 7.72% absolute gain in complex background deepfake detection.

  2. Escaping the Linearity Trap: Manifold Detours for Black-Box Adversarial Attacks on Singing Audio Deepfake Detection

    cs.CR 2026-05 unverdicted novelty 7.0

    MARS is a transfer-based black-box attack that uses bi-level optimization on semantic and artifact anchors to escape the linearity trap and improve attack success rates on SSL-SVDD by up to 36%.

  3. Profiling the Voice: Speaker-Specific Phoneme Fingerprinting for Speech Deepfake Detection

    cs.SD 2026-05 unverdicted novelty 7.0

    PVP models speaker-specific phoneme acoustic distributions with lightweight GMMs trained only on real speech to detect deepfakes of persons-of-interest, outperforming generic detectors and introducing a new Chinese PO...

  4. Benchmarking Multilingual Speech Models on Pashto: Zero-Shot ASR, Script Failure, and Cross-Domain Evaluation

    cs.CL 2026-04 conditional novelty 7.0

    Multilingual ASR models show 39.7-297% zero-shot WER on Pashto public data, Whisper models output correct script in under 0.8% of cases, and fine-tuned models degrade to 32.5-59% WER on out-of-domain sets.

  5. A SUPERB-Style Benchmark of Self-Supervised Speech Models for Audio Deepfake Detection

    eess.AS 2026-03 unverdicted novelty 7.0

    Spoof-SUPERB benchmark shows large-scale discriminative SSL models such as XLS-R, UniSpeech-SAT, and WavLM Large outperform others in audio deepfake detection and maintain robustness under acoustic degradations.

  6. Syntactic Belief Update as the Driver of Garden Path Processing Difficulty

    cs.CL 2026-06 unverdicted novelty 6.0

    Syntactic belief update via generalized Rényi divergence on syntactic trees predicts garden path reading times better than lexical surprisal.

  7. What Do Deepfake Benchmarks Measure? An Audit Using Frozen Self-Supervised Representations

    cs.CV 2026-06 unverdicted novelty 6.0

    Linear probes on frozen self-supervised representations closely approach bespoke deepfake detector performance on benchmarks, indicating benchmarks largely measure general modality understanding.

  8. Dual-Branch Gated Fusion for Open-Set Audio Deepfake Source Tracing

    cs.SD 2026-06 unverdicted novelty 6.0

    A gated fusion of XLSR-53 and CORES features with energy margin and diversity losses reaches 97.6% ID accuracy and reduces FPR95 by 83.5% relative to the Interspeech 2025 baseline on MLAAD.

  9. Forensic Similarity for Speech Deepfakes

    cs.SD 2025-10 unverdicted novelty 6.0

    Introduces forensic similarity for speech deepfakes via a Siamese feature extractor and similarity network to verify shared forensic traces and source models between audio segments.

  10. Impact Analysis of Speech Representation Learning Models for Acoustic Side-Channel Attack

    cs.CR 2026-06 unverdicted novelty 5.0

    KEYAC dataset benchmarks speech models for keyboard acoustic side-channel attacks, with KAN fine-tuning setting new SOTA by addressing nonlinear feature interactions.

  11. Impact Analysis of Speech Representation Learning Models for Acoustic Side-Channel Attack

    cs.CR 2026-06 unverdicted novelty 5.0

    KEYAC dataset created; KAN fine-tuning achieves SOTA on acoustic side-channel keystroke recognition from speech representations under zero-shot and partial fine-tuning.

  12. Light-weight Pronunciation Assessment via Discrete Speech Token Surprisal

    cs.CL 2026-06 unverdicted novelty 5.0

    A framework using native-only trained discrete token surprisal and DTW alignment features improves pronunciation assessment PCC to 0.66 on SpeechOcean762, approaching supervised performance.

  13. Detecting Audio Deepfakes on the Edge:Lightweight SSL-Based Detection in a Browser Plugin

    eess.AS 2026-06 unverdicted novelty 4.0

    Truncated SSL backbone with logistic classifier detects audio deepfakes on-device, claimed to outperform AASIST by 10% while running 40% faster, packaged as a browser plugin.

  14. From Speech to Text Corpora: Evaluating ASR-Based Data Acquisition for Low-Resource Fongbe and Hausa

    cs.CL 2026-06 unverdicted novelty 4.0

    Fine-tuned MMS model reaches 9.48% WER on Fongbe benchmark while Whisper on Hausa videos yields 6,770 segments rated 57.4/100 quality, with Fongbe lower at 36.5/100.

  15. Beyond Speaker Independence: Evaluating Cross-Lingual Acoustic-to-Articulatory Inversion Across Finnish and Russian

    eess.AS 2026-06 unverdicted novelty 4.0

    Benchmarks on the new FROST-EMA corpus show cross-language mismatch drops Pearson correlation by 0.10-0.20 while cross-gender mismatch drops it by 0.05-0.10.

  16. Pretrained self-supervised speech models can recognize unseen consonants

    cs.CL 2026-06 unverdicted novelty 4.0

    Fine-tuned Wav2Vec2 and HuBERT models recognize click consonants more accurately than non-clicks in G|ui and West !Xoon data.

  17. Similarity Choice and Negative Scaling in Supervised Contrastive Learning for Deepfake Audio Detection

    eess.AS 2026-04 unverdicted novelty 4.0

    Cosine similarity in SupCon with a delayed negative queue on wav2vec2 XLS-R yields the lowest equal error rates for deepfake audio detection on in-the-wild and pooled evaluations.

  18. Giving Voice to the Constitution: Low-Resource Text-to-Speech for Quechua and Spanish Using a Bilingual Legal Corpus

    cs.CL 2026-04 unverdicted novelty 4.0

    A bilingual TTS system for the Peruvian Constitution in Quechua and Spanish is developed with XTTS v2, F5-TTS, and DiFlow-TTS, releasing checkpoints and audio to support low-resource speech synthesis.

  19. From Objectives to Applications: Aligning Architectural Biases in Audio Self-Supervised Learning

    eess.AS 2026-07 unverdicted novelty 3.0

    A survey that organizes audio SSL into five objective paradigms, relates their demands to architectural biases, and interprets downstream applications as tests of generalization.

  20. Responsible ASR: Overcoming Challenges of Foundational Models in Narrow-Band and Low-Resource Settings

    cs.SD 2026-06 unverdicted novelty 3.0

    Evaluation of open-source and commercial ASR models on narrow-band Hindi and Indian English shows poor zero-shot results and inconsistent fine-tuning benefits tied to pretraining exposure.

  21. SpAArSIST: Sparsified AASIST for Efficient and Reliable Anti-Spoofing

    cs.SD 2026-06 conditional novelty 3.0

    SpAArSIST sparsifies AASIST by swapping learned pooling for explicit magnitude-based scoring and mean aggregation, cutting compute 20.7% and improving In-the-Wild EER to 2.82%.