List your company
ResearchFormerly WizWand

Speech-to-Text Translation on FLEURS zero-shot evaluation languages (test)

20.84Average BLEUPOTSA (Our Model)
11.157613.671316.18518.6987Nov 12, 2025
Updated 7mo ago

Evaluation results

MethodLinks
POTSA (Our Model)evaluation_mode=zero-shot, Bias Compensation=true, OT Alignment=true2025.1120.84
POTSA (w/o Bias Comp.)evaluation_mode=zero-shot, Bias Compensation=false2025.1119.34
POTSA (w/o OT Align.)evaluation_mode=zero-shot, OT Alignment=false2025.1118.51
POTSA (En-Anchor Fixed)evaluation_mode=zero-shot, Anchor Strategy=Fixed2025.1118.37
WhisperV3+Qwen2.5Training Approach=end-to-end, evaluation_mode=zero-shot2025.1117.91
Qwen2.5-Omnievaluation_mode=zero-shot2025.1116.91
Qwen2-Audioevaluation_mode=zero-shot2025.1116.67
POTSA (En-Anchor Train.)evaluation_mode=zero-shot, Anchor Strategy=Trainable2025.1111.53