Mitigating Label Noise using Prompt-Based Hyperbolic Meta-Learning in Open-Set Domain Generalization

Puntos clave

• The approach significantly reduces label noise, enhancing model reliability in varied environments.
• Meta-learning techniques achieve superior performance with hyperbolic space representations, facilitating better adaptation.
• Analysis employs prompt-based strategies to inform model training and reduce label ambiguity.
• These findings support broader application scenarios, though validation across more datasets is warranted.