Self-regulating Prompts: Foundational Model Adaptation without Forgetting

Benchmark Model Rank Results
prompt-engineering-on-caltech-101PromptSRC#12Harmonic mean: 96.02
prompt-engineering-on-dtdPromptSRC#8Harmonic mean: 71.75
prompt-engineering-on-eurosatPromptSRC#11Harmonic mean: 82.32
prompt-engineering-on-fgvc-aircraftPromptSRC#8Harmonic mean: 40.15
prompt-engineering-on-food-101PromptSRC#7Harmonic mean: 91.10
prompt-engineering-on-imagenetPromptSRC#11Harmonic mean: 74.01
prompt-engineering-on-imagenet-aPromptSRC#5Top-1 accuracy %: 50.90
prompt-engineering-on-imagenet-rPromptSRC#2Top-1 accuracy %: 77.80
prompt-engineering-on-imagenet-sPromptSRC#2Top-1 accuracy %: 49.55
prompt-engineering-on-imagenet-v2PromptSRC#4Top-1 accuracy %: 64.35
prompt-engineering-on-oxford-102-flowerPromptSRC#7Harmonic mean: 85.95
prompt-engineering-on-oxford-iiit-pet-datasetPromptSRC#11Harmonic mean: 96.30
prompt-engineering-on-stanford-cars-1PromptSRC#6Harmonic mean: 76.58
prompt-engineering-on-sun397PromptSRC#10Harmonic mean: 80.52
prompt-engineering-on-ucf101PromptSRC#8Harmonic mean: 82.74