arXiv ScienceSearch

EXPLORE CONNECTIONS

Top-Down Compression: Revisit Efficient Vision Token Projection for Visual Instruction Tuning

Follow the relationships that help you find your next source.

Connections are not prepared for this record yet, or its source metadata has changed. The original record remains available while background snapshots are rebuilt.