RMAdapter: Reconstruction-based Multi-Modal Adapter for Vision-Language Models

Benchmark Model Rank Results
prompt-engineering-on-caltech-101RMAdapter–Harmonic mean: 96.26
prompt-engineering-on-dtdRMAdapter–Harmonic mean: 74.09
prompt-engineering-on-eurosatRMAdapter–Harmonic mean: 84.69
prompt-engineering-on-fgvc-aircraftRMAdapter–Harmonic mean: 39.76
prompt-engineering-on-food-101RMAdapter–Harmonic mean: 91.10
prompt-engineering-on-imagenetRMAdapter–Harmonic mean: 74.52
prompt-engineering-on-imagenet-aRMAdapter–Top-1 accuracy %: 51.21
prompt-engineering-on-imagenet-rRMAdapter–Top-1 accuracy %: 77.70
prompt-engineering-on-imagenet-sRMAdapter–Top-1 accuracy %: 49.50
prompt-engineering-on-imagenet-v2RMAdapter–Top-1 accuracy %: 64.45
prompt-engineering-on-oxford-iiit-pet-datasetRMAdapter–Harmonic mean: 96.89
prompt-engineering-on-stanford-carsRMAdapter–Harmonic mean: 77.75
prompt-engineering-on-ucf101RMAdapter–Harmonic mean: 83.63