Adapting Multi-modal Large Language Model to Concept Drift From Pre-training Onwards

Xiaoyu Yang, Jie Lu, En Yu

Multi-modal Large Language Models (MLLMs) frequently face challenges from concept drift when dealing with real-world streaming data, wherein distributions change unpredictably. Thi…