List your company

Vocal Sound Classification on VocalSound

94.85AccuracyKimi Audio
47.63459.89272.1584.408May 19, 2023Nov 23, 2023May 30, 2024Dec 5, 2024Jun 12, 2025Dec 18, 2025Jun 25, 2026

Evaluation results

MethodLinks
Kimi Audio2026.0294.85
Concat.Encoder Type=Fusion2026.0694.5
Qwen2-Audio2025.0293.9
Adapt. and Trans.Encoder Type=Fusion2026.0693.8
WQ-FusionEncoder Type=Fusion2026.0693.8
Gated Trans.Encoder Type=Fusion2026.0693.6
Qwen2-Audio-7BEncoder Type=Single2026.0693
Qwen-Audiozero-shot=true2023.1192.89
LongCat-Flash-Omni2026.0292.76
Dasheng 1.2BParameter Count=1.2B2026.0192.5
Gemini-3-Pro2026.0292.01
Qwen3-Omni-Instruct2026.0291.6
Whisper-LargeEncoder Type=Single2026.0691.6
ERNIE 5.0-Base2026.0291.48
Challenge BaselineDescription=Best among Dasheng-base, data2vec, and Whisper2026.0191
Ensemble (Submission)Ensemble Strategy=Embedding Concatenation2026.0190.9
AudioMAEEncoder Type=Single2026.0690.9
ERNIE 5.02026.0290.73
Soundwave2025.0290.5
BEATs (300M) SpeechParameter Count=300M, Pre-training Mixture=Speech-heavy (70:15:15)2026.0187.9
BEATs (300M) BalancedParameter Count=300M, Pre-training Mixture=Balanced (40:30:30)2026.0187.7
BEATs (90M) iter3Parameter Count=90M, Iteration=32026.0186.5
Dasheng-BaseEncoder Type=Single2026.0685.5
PALM + ASPLShots=16, Audio Encoder=CLAP-HTSAT, Text Encoder=CLIP (PENGI)2026.0683.03
PALMShots=16, Audio Encoder=CLAP-HTSAT, Text Encoder=CLIP (PENGI)2026.0682.64
PALM + ASPL*Shots=16, Audio Encoder=CLAP-HTSAT, Text Encoder=CLIP (PENGI)2026.0682.6
GPT-4o-Audio2026.0282.37
CoCoOp + ASPLShots=16, Audio Encoder=CLAP-HTSAT, Text Encoder=CLIP (PENGI)2026.0682.27
CoCoOp + ASPL*Shots=16, Audio Encoder=CLAP-HTSAT, Text Encoder=CLIP (PENGI)2026.0682.13
CoCoOpShots=16, Audio Encoder=CLAP-HTSAT, Text Encoder=CLIP (PENGI)2026.0681.17
CoOp + ASPL*Shots=16, Audio Encoder=CLAP-HTSAT, Text Encoder=CLIP (PENGI)2026.0677.28
CoOpShots=16, Audio Encoder=CLAP-HTSAT, Text Encoder=CLIP (PENGI)2026.0674.57
CoOp + ASPLShots=16, Audio Encoder=CLAP-HTSAT, Text Encoder=CLIP (PENGI)2026.0674.11
Pengi2025.0260.4
PengiTask-specific fine-tuning=true2023.1160.35
Pengi2023.0560.35
CLAPTask-specific fine-tuning=true2023.1149.45
CLAP2023.0549.45