Xuri Ge

Smaller multimodal emotion models can outperform their larger counterparts while being more efficient, challenging the prevailing belief that bigger is always better.

Stream-aware adaptation can significantly enhance the performance of large multimodal models in sequential recommendation, overcoming traditional limitations of layer degradation.