Reposted by Lamberto Ballan
What VGGT Knows About Overlap: Probing Geometric Foundation Models for Co-Visibility
Filippo Ziliotto, Luciano Serafini, @lambertoballan.bsky.social, Tommaso Campari
tl;dr: early layers->3D-aware scene representation; late layers->co-visibility; VGGT+MoE head
arxiv.org/abs/2607.09503
