Everything Few-shot

SCHK-HTC: Sibling Contrastive Learning with Hierarchical Knowledge-Aware Prompt Tuning for Hierarchical Text Classification ↗

SCHK-HTC improves few-shot hierarchical text classification by using sibling contrastive learning to distinguish semantically similar classes at deep hierarchy levels, rather than only enforcing parent-child consistency. The method addresses the bottleneck of insufficient domain knowledge for differentiating sibling classes under data-scarce conditions.