AISTAT Lab Advances Language-Based Audio Retrieval for DCASE 2025

AISTAT's dual‑encoder system for DCASE 2025 Task 6, using contrastive learning, achieved a mean average precision@16 of 46.62, rising to 48.83 with a four‑model ensemble. Read more: getnews.me/aistat-lab-advances-lang... #dcase2025 #aistat