SemPT: Semantic Prompt Tuning for Vision-Language Models

Benchmark Model Rank Results
prompt-engineering-on-imagenet-aPromptKD w/ SemPT–Top-1 accuracy %: 70.83
prompt-engineering-on-imagenet-rPromptKD w/ SemPT–Top-1 accuracy %: 87.42
prompt-engineering-on-imagenet-sPromptKD w/ SemPT–Top-1 accuracy %: 59.48
prompt-engineering-on-imagenet-v2PromptKD w/ SemPT–Top-1 accuracy %: 70.46