HPT++: Hierarchically Prompting Vision-Language Models with Multi-Granularity Knowledge Generation and Improved Structure Modeling

Open paper
Benchmark Model Rank Results
prompt-engineering-on-caltech-101HPT++#4Harmonic mean: 96.96
prompt-engineering-on-imagenet-v2HPT++#4Top-1 accuracy %: 65.31
prompt-engineering-on-ucf101HPT++#4Harmonic mean: 83.81
prompt-engineering-on-oxford-iiit-pet-datasetHPT++#5Harmonic mean: 96.91
prompt-engineering-on-dtdHPT++#7Harmonic mean: 74.23
prompt-engineering-on-fgvc-aircraftHPT++#7Harmonic mean: 41.33
prompt-engineering-on-imagenet-aHPT++#7Top-1 accuracy %: 51.18
prompt-engineering-on-sun397HPT++#8Harmonic mean: 81.11
prompt-engineering-on-eurosatHPT++#9Harmonic mean: 87.36
prompt-engineering-on-imagenetHPT++#10Harmonic mean: 74.24
prompt-engineering-on-imagenet-rHPT++#10Top-1 accuracy %: 77.52
prompt-engineering-on-food-101HPT++#11Harmonic mean: 91.09
prompt-engineering-on-oxford-102-flowerHPT++#12Harmonic mean: 85.85
prompt-engineering-on-imagenet-sHPT++#15Top-1 accuracy %: 49.28
prompt-engineering-on-stanford-carsHPT++#16Harmonic mean: 75.59

Rank counts only results with a code link.