New HopS method improves prompt learning for vision-language models with partial labels

Researchers have developed a new method called Holistic Optimal Label Selection (HopS) to improve prompt learning for vision-language models when only partial labels are available. HopS employs two strategies: a local filter that identifies the most plausible label based on nearest neighbors and their softmax scores, and a global objective using optimal transport to map sampling distributions to candidate label distributions. Experiments on eight benchmark datasets demonstrate that HopS consistently enhances performance under partial supervision, outperforming existing methods and offering a practical solution for weakly supervised settings. AI

IMPACT Enhances prompt learning for vision-language models in weakly supervised scenarios, potentially improving their adaptability to diverse datasets.

RANK_REASON The cluster contains a research paper detailing a new method for prompt learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• benchmark dataset
• feature encoders
• Haoliang Sun
• Holistic Optimal Label Selection
• Hugging Face
• optimal transport
• softmax scores
• vision-language model

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →