A large-scale, difficulty-stratified benchmark for temporal grounding of sound events
in audio, introduced in the paper Auto-AEG: Scalable Data Construction for
Open-Vocabulary Audio Event Grounding.
Given an audio clip and a natural-language query (e.g. "growling"), a model must
output the [onset, offset] time intervals in which the queried event occurs, or
return [] (empty) when the event is absent.
AEGBench evaluates two… See the full description on the dataset page:
https://huggingface.co/datasets/zihan-audio/AEGBench.
