AlphaWiSE: Adaptive Weight Interpolation for Continual Multimodal Representation Learning — ThinkLLM