Google FLEURS — Alexis Conneau, Min Ma, Simran Khanuja, Yu Zhang, Vera Axelrod, Siddharth Dalmia, Jason Riesa, Clara Rivera, Ankur Bapna (2022). FLEURS: Few-shot Learning Evaluation of Universal Representations of Speech.
https://arxiv.org/abs/2205.12446
https://huggingface.co/datasets/google/fleurs
Revision: 70bb2e84b976b7e960aa89f1c648e09c59f894dd; test split.
CC BY 4.0: https://creativecommons.org/licenses/by/4.0/
Legal code: https://creativecommons.org/licenses/by/4.0/legalcode
Original WAV and independent reference texts unchanged. Automatic SRTs are Subvideo product outputs dated 2026-09-29, not corrected references. Derived diagnoses: synthetic zero-sample silence; repeated ja-01 separated by 5 seconds silence; ja-01 and de-01 with Gaussian noise, seed 20260929, 10 dB SNR, whole-signal gain 1.0, PCM16. Parameters: diagnostics.json.
Presentation, selection and metrics by Subvideo; no endorsement, official FLEURS ranking or independent certification. Original results retained, including historical silence placeholder fixed in the product on 2026-09-30.