Connect, Collapse, Corrupt: Learning Cross-Modal Tasks with Uni-Modal Data

Open in new window