ChildVox: A Speech, Audio, and Large Audio-Language Model Benchmark in Understanding and Characterizing Sound across Childhood

Researchers introduced ChildVox, a benchmark for child speech and audio understanding from birth through school age.

It combines 20+ tasks from 17 child-focused datasets to evaluate self-supervised, ASR, and large audio-language models.

The benchmark measures how well models handle child vocalizations, speech recognition, and related acoustic signals.

ChildVox provides a standardized way to compare model performance on child-specific audio across developmental stages.