1
GRID: Scaling Task-Agnostic Inference in Continual Prompt Tuning
提出GRID框架，实现连续提示调优中任务无关推理的规模化，解决增量学习场景下的模型适应难题，理论创新与实验验证兼具。
arXiv:2507.14725v4 Announce Type: replace-cross Abstract: Prompt-based continual learning (CL) offers a parameter-efficient way to adapt large languag…