IMP-MoE-L

Reported on 4 benchmarks across 2 tasks · 1 paper · 2 SOTA

Note: results are matched by exact model name. Different papers may use the same name for different model variants.

Computer Vision4 results

Zero-Shot Action RecognitiononUCF101
Top-1 Accuracy· uses extra data· 2023-05-10
91.5
best: 92.8 (OTI(ViT-L/14))
SOTA
Alternating Gradient Descent and Mixture-of-Experts for Integrated Multimodal Perception arXiv:2305.06324
Zero-Shot Action RecognitiononKinetics
Top-1 Accuracy· uses extra data· 2023-05-10
76.8
best: 78.1 (TC-CLIP)
SOTA
Alternating Gradient Descent and Mixture-of-Experts for Integrated Multimodal Perception arXiv:2305.06324
Zero-Shot Transfer Image ClassificationonImageNet
Accuracy (Private)· 2023-05-10
83.9
best: 88.5 (M2-Encoder)
Alternating Gradient Descent and Mixture-of-Experts for Integrated Multimodal Perception arXiv:2305.06324
Zero-Shot Action RecognitiononHMDB51
Top-1 Accuracy· uses extra data· 2023-05-10
59.1
best: 64.7 (MOV (ViT-L/14))
Alternating Gradient Descent and Mixture-of-Experts for Integrated Multimodal Perception arXiv:2305.06324