List your company
ResearchFormerly WizWand

Long speech understanding and reasoning on MMBench Audio

0.4938AccuracyUni-MoE
-0.0004080.1278960.25620.384504May 18, 2024
Updated 8mo ago

Evaluation results

MethodLinks
Uni-MoETraining Task=MoE-Task2, Training Epochs=2 epoch2024.050.4938
Single-Modality-Expert-Task7Evaluation Protocol=zero-shot2024.050.4783
Single-Modality-Expert-Task6Evaluation Protocol=zero-shot2024.050.4752
Uni-MoETraining Task=MoE-Task1, Training Epochs=3 epoch2024.050.472
Single-Modality-Expert-Task4Evaluation Protocol=zero-shot2024.050.4596
Uni-MoETraining Task=MoE-Task1, Training Epochs=2 epoch2024.050.4441
Single-Modality-Expert-Task5Evaluation Protocol=zero-shot2024.050.413
Uni-MoETraining Task=MoE-Task1-short-speech, Training Epochs=2 epoch2024.050.3758
Uni-MoETraining Task=MoE-Task1-short-speech, Training Epochs=3 epoch2024.050.3478
X-InstructBLIPEvaluation Protocol=zero-shot2024.050.1123
Macaw-LLMEvaluation Protocol=zero-shot2024.050.0186