List your company

Automatic Speech Recognition on FLEURS South-Asia (14 languages from the 8th Schedule) (test)

0.5338WERSigLIP2-base
0.528040.566920.60580.64468Jun 23, 2026

Evaluation results

MethodLinks
SigLIP2-baseInput configuration=single-image2026.060.533813131310
SigLIP2-largeInput configuration=multi-image2026.060.535813131311
Qwen3-VLInput configuration=multi-image2026.060.568312101022
BaselineConfiguration=SSL only2026.060.6778—————