List your company

Audio-to-Text Retrieval on FSD50K

69.7R@1Proposed
53.5857.76561.9566.135Apr 25, 2026

Evaluation results

MethodLinks
ProposedModality=a2t, audio encoder=HTSAT-tiny, text encoder=RoBERTa-large2026.0469.788.992.967.2
LAION-CLAPModality=a2t2026.0465.587.491.365.9
Microsoft-CLAPModality=a2t2026.0454.283.789.758.1