audio benchmark
VocalSound
A dataset for improving human vocal sounds recognition, containing over 21,000 crowdsourced recordings of laughter, sighs, coughs, throat clearing, sneezes, and sniffs from 3,365 unique subjects. Used for audio event classification and recognition of human non-speech vocalizations.
Updated Aug 7, 2026
Published models1
Registry coverage1
MetricScore
EvidenceB
VocalSound leaderboard
Sorted by the source-provided rank. Higher score is better according to the registry.
| 1 | AC | 93.9% | 100.0% | 1 | C |
VocalSound highlights
The top published results on this benchmark's own scale.
What is VocalSound?
Definition and scoring fields from the benchmark registry.
A dataset for improving human vocal sounds recognition, containing over 21,000 crowdsourced recordings of laughter, sighs, coughs, throat clearing, sneezes, and sniffs from 3,365 unique subjects. Used for audio event classification and recognition of human non-speech vocalizations.
Scores are shown in ratio. The current registry marks this benchmark as not independently verified with evidence level B.
- Family
- VocalSound
- Modality
- audio
- Primary category
- audio
- Score direction
- higher
- LLMBoard eligible
- No
- Evaluation key
- vocalsound|llm-stats-current
Source-native results are preserved. Eligibility for the overall LLMBoard score is a separate policy decision.
FAQ
Common questions about VocalSound.
Which model scores highest on VocalSound?
Qwen2.5-Omni-7B is currently ranked first with 93.9%.
What does VocalSound measure?
A dataset for improving human vocal sounds recognition, containing over 21,000 crowdsourced recordings of laughter, sighs, coughs, throat clearing, sneezes, and sniffs from 3,365 unique subjects. Used for audio event classification and recognition of human non-speech vocalizations.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
1 unique published model results are currently shown.
Does this benchmark affect the overall score?
This benchmark is preserved as source-native evidence but is not eligible for the current overall score.