In-Context Learning Framework for Zero-Shot Extreme Multi-Label Classification

A two-stage framework that leverages in-context learning to generate and rerank candidate labels for zero-shot extreme multi-label classification tasks.