List your company
ResearchFormerly WizWand

Audio-Visual Question Answering on Music-AVQA

81.3AccuracyVideoLLaMA2
25.795240.205154.61569.0249Apr 17, 2023Oct 29, 2023May 12, 2024Nov 24, 2024Jun 7, 2025Dec 20, 2025Jul 4, 2026
Updated 3mo ago

Evaluation results

MethodLinks
VideoLLaMA2FLOPs=100, Latency=0.43, Memory=22G2026.0181.3
VideoLLaMA2 w/ FastAVFLOPs=56, Latency=0.32, Memory=19G2026.0181.2
VALORLPretraining examples=33.5M, Modality=V+A2023.0478.9
Q-TriM2026.0777.68
VALORBPretraining examples=6.5M, Modality=V+A2023.0476.6
MAVEN2026.0776.21
VALORLPretraining examples=5.5M, Modality=V2023.0474.8
MUSIC-AVQAModality=V+A2023.0471.5
VideoLLaMA22026.0766.85
Qwen2.5-VLZero-shot=true2026.0757.07
GPT-4oZero-shot=true2026.0756.06
TSV MergingEvaluation Protocol=Zero-shot, Composition Strategy=TSV Merging2025.0553.78
MMER2025.0553.54
NaiveMCEvaluation Protocol=Zero-shot, Composition Strategy=NaiveMC2025.0553.5
OptMergeEvaluation Protocol=Zero-shot, Composition Strategy=OptMerge2025.0553.17
DAMCEvaluation Protocol=Zero-shot, Composition Strategy=DAMC2025.0552.8
Iso-CEvaluation Protocol=Zero-shot, Composition Strategy=Iso-C2025.0552.77
WUDI MergingEvaluation Protocol=Zero-shot, Composition Strategy=WUDI Merging2025.0552.43
Task ArithmeticEvaluation Protocol=Zero-shot, Composition Strategy=Task Arithmetic2025.0552.14
VisionEvaluation Protocol=Zero-shot, Input Modality=Vision2025.0550.77
Ties MergingEvaluation Protocol=Zero-shot, Composition Strategy=Ties Merging2025.0550.35
AVicunaModel size=7B, # Pairs=1.1M, Zero-shot=true2024.1149.6
LongVALE-LLMModel size=7B, # Pairs=0.7M, Zero-shot=true2024.1149.4
VideoEvaluation Protocol=Zero-shot, Input Modality=Video2025.0549.02
CAT-7Bzero-shot=true2024.0348.6
Weight AverageEvaluation Protocol=Zero-shot, Composition Strategy=Weight Average2025.0547.75
Video MLLMModality=Video2025.0547.72
ChatBridge-13Bzero-shot=true2024.0347.6
OneLLMModel size=7B, # Pairs=1007M, Zero-shot=true2024.1147.6
AV-LLMModel size=13B, # Pairs=1.6M, Zero-shot=true2024.1145.2
X-InstructBLIPModel size=13B, # Pairs=32M, Zero-shot=true2024.1144.5
Vision MLLMModality=Vision2025.0544.06
OneLLM-7Bzero-shot=true2024.0343
VideoLLaMAModel size=7B, # Pairs=2.8M, Zero-shot=true2024.1136.6
PandaGPTModel size=13B, # Pairs=128M, Zero-shot=true2024.1133.7
Macaw-LLMModel size=7B, # Pairs=0.3M, Zero-shot=true2024.1131.8
Audio MLLMModality=Audio2025.0530.63
AudioEvaluation Protocol=Zero-shot, Input Modality=Audio2025.0527.93