Self-regulating Prompts: Foundational Model Adaptation without Forgetting

Benchmark Model Rank Results
prompt-engineering-on-imagenet-rPromptSRC#7Top-1 accuracy %: 77.80
prompt-engineering-on-food-101PromptSRC#8Harmonic mean: 91.10
prompt-engineering-on-imagenet-sPromptSRC#8Top-1 accuracy %: 49.55
prompt-engineering-on-imagenet-aPromptSRC#9Top-1 accuracy %: 50.90
prompt-engineering-on-ucf101PromptSRC#9Harmonic mean: 82.74
prompt-engineering-on-oxford-102-flowerPromptSRC#11Harmonic mean: 85.95
prompt-engineering-on-stanford-carsPromptSRC#11Harmonic mean: 76.58
prompt-engineering-on-fgvc-aircraftPromptSRC#13Harmonic mean: 40.15
prompt-engineering-on-imagenet-v2PromptSRC#13Top-1 accuracy %: 64.35
prompt-engineering-on-dtdPromptSRC#15Harmonic mean: 71.75
prompt-engineering-on-imagenetPromptSRC#15Harmonic mean: 74.01
prompt-engineering-on-sun397PromptSRC#16Harmonic mean: 80.52
prompt-engineering-on-eurosatPromptSRC#20Harmonic mean: 82.32
prompt-engineering-on-oxford-iiit-pet-datasetPromptSRC#23Harmonic mean: 96.30
prompt-engineering-on-caltech-101PromptSRC#25Harmonic mean: 96.02