Prompt Based Pseudo

Prompt-based pseudo-labeling leverages the power of large language models (LLMs) and vision-language models (VLMs) to generate or refine labels for data, particularly in scenarios with limited labeled examples. Current research focuses on automating prompt generation and optimization, often employing techniques like prompt tuning and iterative relabeling, to improve the accuracy and efficiency of pseudo-label creation across various tasks including image classification, text summarization, and OOD detection. This approach holds significant promise for advancing semi-supervised learning and reducing the reliance on expensive human annotation in diverse fields, ultimately improving the performance and scalability of machine learning models.