Task-Adaptive Parameter-Efficient Fine-Tuning for Weather Foundation Models
Researchers have developed WeatherPEFT, a new parameter-efficient fine-tuning framework specifically designed for weather foundation models. This framework addresses the unique challenges of weather-related tasks, such as variable heterogeneity and resolution diversity, which standard PEFT methods struggle with. WeatherPEFT incorporates Task-Adaptive Dynamic Prompting to recalibrate features contextually and Stochastic Fisher-Guided Adaptive Selection to identify critical parameters, achieving performance comparable to full fine-tuning with fewer trainable parameters. AI
IMPACT This research offers a more efficient way to adapt large weather models for specific tasks, potentially lowering deployment barriers.