List your company

Audio-visual Speech-to-Text Translation on MuAViC (test)

21.7BLEU (EL->EN)XLAVS-R 2B
-0.7645.06810.916.732Mar 21, 2024Jul 4, 2024Oct 18, 2024Feb 1, 2025May 18, 2025Sep 1, 2025Dec 16, 2025

Evaluation results

MethodLinks
XLAVS-R 2BTest Mode=A, Environment=Clean2024.0321.72530.626.530.213.924.7
XLAVS-R 2BTest Mode=AV, Environment=Clean2024.0321.625.130.626.629.913.924.6
XLAVS-R 300MTest Mode=AV, Environment=Clean2024.0319.123.829.82528.81323.2
XLAVS-R 300MTest Mode=A, Environment=Clean2024.0318.923.729.724.928.512.623
XLAVS-R 2BTest Mode=AV, Environment=Noisy2024.0315.719.224.620.122.310.418.7
XLAVS-R 2BPretrain data=1.2k hrs, 9 langs, SNR=02025.1215.719.224.620.122.310.418.7
u-HuBERTTest Mode=AV, Environment=Clean2024.0314.523.128.623.72711.421.4
u-HuBERTTest Mode=A, Environment=Clean2024.0314.42328.623.426.611.321.2
AV-HuBERTTest Mode=AV, Environment=Clean2024.0314.322.928.323.926.511.221.2
AV-HuBERTTest Mode=A, Environment=Clean2024.0313.922.328.123.526.110.720.8
XLAVS-R 300MTest Mode=AV, Environment=Noisy2024.0313.217.423.818.721.89.417.4
XLS-R 300MPretrain data=8.3k hrs, 100+ langs, SNR=02025.1213.217.423.818.721.89.417.4
MoHAVE-LargePretrain data=7.0k hrs, 100+ langs, SNR=02025.1211.422.327.122.125.19.219.5
XLAVS-R 2BTest Mode=A, Environment=Noisy2024.031115.120.916.216914.7
mAV-HuBERT (re-implementation)Pretrain data=7.0k hrs, 100+ langs, SNR=02025.128.921.526.521.224.28.818.5
AV-HuBERTTest Mode=AV, Environment=Noisy2024.038.815.619.21517.67.213.9
XLAVS-R 300MTest Mode=A, Environment=Noisy2024.038.313.920.415.215.18.213.5
u-HuBERTTest Mode=AV, Environment=Noisy2024.038.215.619.714.618.37.314
mAV-HuBERT + Hard RoutingPretrain data=7.0k hrs, 100+ langs, SNR=02025.126.719.924.719.6237.216.8
u-HuBERTTest Mode=A, Environment=Noisy2024.034.79.813.88.710.35.38.8
AV-HuBERTTest Mode=A, Environment=Noisy2024.034.49.113.18.38.84.88.1
mAV-HuBERTPretrain data=1.7k hrs, English, SNR=02025.124.212.81512.514.84.610.7
Whisper large-v2Pretrain data=680k hrs, 100+ langs, SNR=02025.120.10.40.70.10.10.20.3