Reproduce “NLPrompt: Noise-Label Prompt Learning for Vision-Language Models” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Bikang Pan, Qun Li, Xiaoying Tang, Wei Huang, Zhen Fang, Feng Liu, Jingya Wang, Jingyi Yu and 1 more
NLPrompt adapts CLIP prompts to noisy labels by using image–text alignment to estimate which examples agree with their supplied labels. It applies cross-entropy to estimated-clean examples and MAE to estimated-noisy examples. On Food101N, it reports 76.46% test accuracy, compared with 72.57% for JoAPR. This result comes from the paper’s reported setup and does not establish performance across real-world noise settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.