MAD: Modality-Adaptive Decoding for Mitigating Cross-Modal Hallucinations in Multimodal Large Language Models

Benchmark Model Rank Results
audio-visual-question-answering-on-music-avqaVideoLLaMA2-AV + MAD#3Acc: 79.1
audio-visual-question-answering-on-omnibenchVideoLLaMA2-AV + MAD#7Accuracy: 36.8
audio-visual-question-answering-on-worldsenseVideoLLaMA2-AV + MAD#17Accuracy: 25.6