Full Parameter Tuning is a model optimization technique in deep learning, especially used in scenarios of transfer learning or domain adaptation. It involves fine-tuning all parameters of a pre-trained model to adapt to a specific task or dataset. This technique allows the model to optimize for a specific task while maintaining pre-trained knowledge, but requires more computing resources accordingly. It is a mainstream paradigm for pre-trained language models (PLMs) in natural language processing (NLP), involving fine-tuning all parameters of the model using annotated data from downstream tasks to adapt to a specific task. Although this method can bring performance improvements, it is also accompanied by huge consumption of computing and storage resources. As the model size increases, the resource requirements of full parameter fine-tuning also increase accordingly, which to some extent limits its scope of application. .