What is the defining purpose of multimodal fusion?

What is the defining purpose of multimodal fusion?

1. Combine representations or decisions from modalities such as text, image, and audio so they inform one task
2. Convert every modality into raw text and discard all non-text structure before encoding
3. Train independent models whose outputs are never compared or combined