AIR-Bench

'AIR-Bench' is a benchmark dataset used to evaluate the performance of audio-text embedding models through multiple-choice questions (MCQs) and zero-shot classification tasks.

Papers using AIR-Bench (3)

WavLink: Compact Audio-Text Embeddings with a Global Whisper Token2026SLM-TTA: A Framework for Test-Time Adaptation of Generative Spoken Language Models2025Soundwave: Less is More for Speech-Text Alignment in LLMs2025