Unbundling Multimodal Vision Encoding
Comments
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules
Summary
The paper "Unbundling Multimodal Vision Encoding" tests the three design choices bundled in a pretrained vision encoder: a visual prior from separate pretraining, dedicated vision capacity, and placement before fusion with text.