Pre-teaching details with a little proportion of multi-task instruction information enhances the overall model general performance In textual unimodal LLMs, textual content will be the unique medium of perception, with other sensory inputs currently being disregarded. This textual content ser