GLAD: Generalizable Tuning for Vision-Language Models

Benchmark Model Rank Results
prompt-engineering-on-caltech-101GLAD–Harmonic mean: 96.49
prompt-engineering-on-dtdGLAD–Harmonic mean: 72.83
prompt-engineering-on-eurosatGLAD–Harmonic mean: 84.22
prompt-engineering-on-fgvc-aircraftGLAD–Harmonic mean: 40.93
prompt-engineering-on-food-101GLAD–Harmonic mean: 91.46
prompt-engineering-on-imagenetGLAD–Harmonic mean: 75.04
prompt-engineering-on-imagenet-aGLAD–Top-1 accuracy %: 51.03
prompt-engineering-on-imagenet-rGLAD–Top-1 accuracy %: 77.43
prompt-engineering-on-imagenet-v2GLAD–Top-1 accuracy %: 64.93
prompt-engineering-on-oxford-102-flowerGLAD–Harmonic mean: 85.35
prompt-engineering-on-oxford-iiit-pet-datasetGLAD–Harmonic mean: 96.78
prompt-engineering-on-stanford-carsGLAD–Harmonic mean: 78.12
prompt-engineering-on-sun397GLAD–Harmonic mean: 81.46
prompt-engineering-on-ucf101GLAD–Harmonic mean: 83.88