Query, Align, and Distill: Navigation-Aware Cross-Modal Interaction for Efficient Vision-and-Language Navigation
A high-performing teacher is built that makes navigation evidence selection explicit and compressible, and a compact student is trained by transferring both where to attend and what to do, then further match action distributions during fine-tuning.
Zhihao Chen, Yi-Yuan Ge, Zi-Yang Wang et al.
· IEEE transactions on circuit... · 0 citations