List your company

Audio Classification on VGG-Sound

88.15Top-1 AccuracyCALM
-3.193220.520944.23567.9491Oct 20, 2022May 16, 2023Dec 11, 2023Jul 7, 2024Feb 1, 2025Aug 29, 2025Mar 26, 2026

Evaluation results

MethodLinks
CALMBackbone=Qwen2-Audio, Pseudo-Labeling=false2026.0288.15—————
CALMBackbone=Qwen2.5-Omni, Pseudo-Labeling=false2026.0287.85—————
CALMBackbone=Qwen2-Audio, Pseudo-Labeling=true2026.0287.67—————
SAVsBackbone=Qwen2.5-Omni, Pseudo-Labeling=false2026.0287.11—————
CALMBackbone=Qwen2.5-Omni, Pseudo-Labeling=true2026.0286.28—————
SAVsBackbone=Qwen2-Audio, Pseudo-Labeling=true2026.0285.65—————
SAVsBackbone=Qwen2-Audio, Pseudo-Labeling=false2026.0285.64—————
SAVsBackbone=Qwen2.5-Omni, Pseudo-Labeling=true2026.0281.78—————
Zero-shotBackbone=Qwen2-Audio2026.0272.94—————
Zero-shotBackbone=Qwen2.5-Omni2026.0261.57—————
ONE-PEACEEvaluation Protocol=Fine-Tuning, Input Modality=Audio Only2023.0559.6—————
SOTA [135]Evaluation Protocol=Fine-Tuning, Input Modality=Audio Only2023.0559.5—————
SAM-2.7Br=256, Conn.=(a), Experiment=E62025.0957—————
SAM-2.7Br=256, Conn.=(b), Ablation=4.2, Experiment=E142025.0956.7—————
SAM-2.7Br=256, Conn.=(c), Ablation=4.2, Experiment=E152025.0956.7—————
SAM-2.7Br=8, Conn.=(a), Experiment=E32025.0956.6—————
MTC (AudioSet)Label=Tag, Evaluation Protocol=linear probing, Pooling Strategy=mean-pooling2026.0356.46—————
SAM-780Mr=256, Conn.=(c), Ablation=4.2, Experiment=E132025.0956.2—————
SAM-780Mr=256, Conn.=(a), Experiment=E52025.0956.1—————
SAM-780Mr=256, Conn.=(b), Ablation=4.2, Experiment=E122025.0955.7—————
SAM-2.7Br=256, Conn.=(a), Ablation=4.3, Experiment=E182025.0955.6—————
LAION-CLAPEvaluation Protocol=Fine-Tuning, Input Modality=Audio Only, reproduced=true2023.0555.1—————
SAM-780Mr=8, Conn.=(a), Experiment=E22025.0954.5—————
SAM-780Mr=256, Conn.=(a), Ablation=4.3, Experiment=E172025.0954.4—————
SAM-2.7Br=256, Conn.=(a), Ablation=4.1, Experiment=E92025.0954.3—————
SAM-130Mr=256, Conn.=(a), Experiment=E42025.0953.5—————
SAM-780Mr=256, Conn.=(a), Ablation=4.1, Experiment=E82025.0953.3—————
SAM-130Mr=256, Conn.=(b), Ablation=4.2, Experiment=E102025.0953.1—————
SAM-130Mr=256, Conn.=(c), Ablation=4.2, Experiment=E112025.0952.9—————
Absolute SOTAReference=[36]2023.0552.5—————
MBT2023.1252.3—————
GAMA-7Br=8, Conn.=QFormer2025.0952.2—————
SAM-130Mr=8, Conn.=(a), Experiment=E12025.0951.8—————
SAM-130Mr=256, Conn.=(a), Ablation=4.3, Experiment=E162025.0951.7—————
ssLALM-2.8Br=8, Conn.=Linear2025.0951—————
LTU-7Br=8, Conn.=Linear2025.0950.3—————
SAM-130Mr=256, Conn.=(a), Ablation=4.1, Experiment=E72025.0948.1—————
UIO-2XXLModel Size=XXL2023.1247.7—————
Wav2CLIPEvaluation Protocol=Fine-Tuning, Input Modality=Audio Only2023.0546.6—————
LAION-CLAPEvaluation Protocol=Zero-shot (No fine-tuning), Augmentation Strategy=keyword-to-caption2023.0446.2—————
UIO-2XLModel Size=XL2023.1244.2—————
Contrastive-scratchLabel=Caption, Evaluation Protocol=linear probing, Pooling Strategy=mean-pooling2026.0343.78—————
B.3Evaluation Protocol=Zero-shot (No fine-tuning), Variant=Monolingual (EN)2023.0443.49—————
B.5Evaluation Protocol=Zero-shot (No fine-tuning), Variant=Multilingual (+EFJ)2023.0442.96—————
BLATEvaluation Protocol=Zero-shot (No fine-tuning)2023.0442.9—————
B.4Evaluation Protocol=Zero-shot (No fine-tuning), Variant=Bilingual2023.0441.95—————
Multi-Task (Ours)Label=Tag*, Caption*, Evaluation Protocol=linear probing, Pooling Strategy=mean-pooling2026.0340.81—————
Contrastive-scratch (Ours)Label=Caption*, Evaluation Protocol=linear probing, Pooling Strategy=mean-pooling2026.0340.78—————
Captioning-scratch (Ours)Label=Caption*, Evaluation Protocol=linear probing, Pooling Strategy=mean-pooling2026.0339.78—————
PAR (Ours-UTS)Label=Tag Sequence*, Evaluation Protocol=linear probing, Pooling Strategy=mean-pooling2026.0339.59—————
Captioning-scratchLabel=Caption, Evaluation Protocol=linear probing, Pooling Strategy=mean-pooling2026.0339.52—————
LanguageBindSize=Large, Evaluation Protocol=Full tuning2023.1038.6—————
UIO-2LModel Size=L2023.1237.8—————
MTC (Ours-UTS)Label=Tag*, Evaluation Protocol=linear probing, Pooling Strategy=mean-pooling2026.0337.7—————
PMRLsetting=zero-shot2025.0736.43—————
GRAMsetting=zero-shot2025.0734.58—————
VASTsetting=zero-shot2025.0733.58—————
WavLink-Smallmode=zero-shot2026.0131.8—————
MGA-CLAPmode=zero-shot2026.0131.8—————
ViT-LENS_Lanchor=I+T2023.1131.7—————
WavLink-Largemode=zero-shot2026.0131.7—————
ImageBindsetting=zero-shot2025.0730.98—————
ReCLAPmode=zero-shot2026.0129.2—————
SOTA [89]2023.1129.1—————
LAION-CLAPmode=zero-shot2026.0129.1—————
LanguageBindSize=Large, Evaluation Protocol=Zero-shot2023.1028.9—————
ViT-LENS_Lanchor=I2023.1128.2—————
ImageBindSize=Huge, Evaluation Protocol=Zero-shot2023.1027.8—————
ImageBind-Hanchor=I2023.1127.8—————
IMAGEBINDZero-shot=true, Text Prompting=true2023.0527.8—————
WavLink-Basemode=zero-shot2026.0127.7—————
AF-CLAPmode=zero-shot2026.0124.1—————
MBT (AV)Modality=Audio-Visual2022.100.6410.856————
PlayItBackX3Modality=Audio2022.100.5370.7920.5610.9782.846—
Slow-FastModality=Audio2022.100.5240.7810.5440.9742.761—
MBT (A)Modality=Audio2022.100.5230.781————
PolyViTModality=Audio-Visual2022.100.517—————
Chen et al.Modality=Audio2022.100.510.7640.5320.9732.735—
Peng et al. (AV)Modality=Audio-Visual2022.100.506—0.539———
ResNet-101Modality=Audio2022.100.4560.7230.4760.9682.615—
Peng et al. (A)Modality=Audio2022.100.443—0.484———
McDonnell & GaoModality=Audio2022.100.3970.7160.4030.9632.532—
RandomZero-shot=true2023.050.32—————
A-RWKV-B/16Training Protocol=from-scratch, Precision=bf162025.09—————45.37
AST-B/16Training Protocol=from-scratch2025.09—————37.25
AST-B/16Training Protocol=from-scratch, Precision=bf162025.09—————39.88
AuM-B/16Training Protocol=from-scratch2025.09—————42.58
CacophonyEvaluation Protocol=Zero-shot, Template=The sound of {}2026.04—————27
FineLAPEvaluation Protocol=Zero-shot, Template=The sound of {}2026.04—————31.5
FLAMEvaluation Protocol=Zero-shot, Template=The sound of {}2026.04—————39.3
HTSAT-BERTEvaluation Protocol=Zero-shot, Template=The sound of {}2026.04—————29.6
LAION-CLAPEvaluation Protocol=Zero-shot, Template=The sound of {}2026.04—————29.1
MGA-CLAPEvaluation Protocol=Zero-shot, Template=The sound of {}2026.04—————31.8